Synthetic media in the digital landscape PDF Free Download
1 / 38/38
100%
Synthetic media
in the digital landscape
A paper by the Canadian Digital Regulators Forum
Canadian Radio-television and Telecommunications Commission
Competition Bureau Canada
Copyright Board of Canada
Office of the Privacy Commissioner of Canada
Canadian Digital Regulators Forum:
Synthetic media in the digital landscape
Aussi disponible en français sous le titre : Forum canadien des organismes de réglementation numérique : Les
médias synthétiques dans le paysage numérique.
For more information, contact
Office of the Privacy Commissioner of Canada
30 Victoria Street
Gatineau, Quebec K1A 1H3
© His Majesty the King in Right of Canada for the Office of the Privacy Commissioner of Canada, 2025
Cat. No.: IP54-119/2025E-PDF
ISBN: 978-0-660-78343-7s
2
Canadian Digital Regulators Forum:
Synthetic media in the digital landscape
Contents
Preamble ...................................................................................................................................... 3
Profiles of participating Members ..................................................................................................... 3
Executive Summary ....................................................................................................................... 4
Purpose .................................................................................................................................... 4
Introduction .................................................................................................................................. 5
What is synthetic media? ............................................................................................................ 5
Synthetic media and the Copyright Board .......................................................................................... 9
Considerations ........................................................................................................................ 11
Final observations .................................................................................................................... 14
Synthetic media and the Canadian Radio-television and Telecommunications Commission .................. 14
Considerations ........................................................................................................................ 15
Final observations .................................................................................................................... 19
Synthetic Media and the Competition Bureau .................................................................................. 19
Considerations ........................................................................................................................ 20
Final observations .................................................................................................................... 24
Synthetic media and the Office of the Privacy Commissioner............................................................. 25
How the federal privacy law for organizations applies to synthetic media......................................... 25
Considerations ........................................................................................................................ 26
Final observations .................................................................................................................... 32
Synthetic Media and Canada’s Anti-Spam Legislation ...................................................................... 33
Commercial electronic messages (Enforced by the CRTC) ............................................................ 33
Deceptive marketing practices (Enforced by the Competition Bureau) ............................................ 34
Conclusion and key takeaways ...................................................................................................... 36
3
Preamble
The Canadian Digital Regulators Forum (CDRF) is a partnership between the Canadian Radio-
television and Telecommunications Commission (CRTC), the Competition Bureau, the Oice of the
Privacy Commissioner of Canada and the Copyright Board of Canada (the Members). The Forum
was established in June 2023 to facilitate information sharing and collaboration on matters that
relate to digital markets.
The CDRF’s inaugural year was dedicated to strengthening the CDRF’s relationships and
collaboration among its Members, as well as its international ties by joining the International
Network for Digital Regulation Cooperation. The CDRF Members also increased their internal
capacities and knowledge of artificial intelligence (AI) through engagement with academics and
industry experts.
In its second year, the Forum’s Members focused their attention on generative AI (gen AI) and
synthetic media (artificial content produced using AI or other automated technologies), which are
becoming central topics of discussion and debate across various industries and sectors.
Profiles of participating Members
About the Copyright Board
The Copyright Board is an independent administrative tribunal and economic regulator that grants
licences and establishes fair and equitable taris for the use of protected works. The Board plays
an essential role in the copyright marketplace, balancing the remuneration of copyright owners with
providing user access to works, while preserving public interest and market competitiveness.
To deliver on its mandate, the Board is required to act fairly, base its work on solid legal and
economic principles, and reflect a firm understanding of evolving business models and
technologies.
As an independent tribunal, the Board reports on its administrative activities to Parliament through
the Minister of Innovation, Science and Industry.
About the Canadian Radio-television and Telecommunications Commission
The Canadian Radio-television and Telecommunications Commission (CRTC) is an independent
quasi-judicial tribunal that regulates the Canadian communications sector in the public interest.
The CRTC holds public consultations on telecommunications and broadcasting matters and makes
decisions based on the public record. The CRTC has a wide range of responsibilities, including
under the Telecommunications Act, the Broadcasting Act, the Online News Act, and Canada’s Anti-
Spam Legislation (CASL).
The CRTC currently has nine members, including a Chairperson, a Vice-Chairperson for
Telecommunications, a Vice-Chairperson for Broadcasting, and six regional Commissioners who
are located across the country. Each is an expert in their field and brings a unique perspective to the
CRTC’s work. Supporting the nine decision-makers is a team of expert staff.
4
About the Competition Bureau
Headed by the Commissioner of Competition, the Competition Bureau is an independent law
enforcement agency that protects and promotes competition for the benefit of Canadian
consumers and businesses. Among other laws, the Competition Bureau is responsible for
administering and enforcing the Competition Act. The Competition Act contains both criminal and
civil provisions aimed at preventing anti-competitive practices in the marketplace.
Competition drives lower prices and innovation while fueling economic growth. See the
Competition Bureau’s page on why competition matters.
About the Oice of the Privacy Commissioner of Canada
The Privacy Commissioner of Canada is an Agent of Parliament whose mission is to protect and
promote privacy rights. The Oice of the Privacy Commissioner of Canada (OPC) oversees
compliance with the Privacy Act, which covers the personal information-handling practices of
federal government departments and agencies, and the Personal Information Protection and
Electronic Documents Act (PIPEDA), Canada’s federal private-sector privacy law.
The mission of the OPC is to protect and promote the privacy rights of individuals. The Privacy
Commissioner of Canada, who is independent of government, reports directly to Parliament.
Executive Summary
Purpose
The CDRF recognizes that synthetic media has significant implications for Canadians and
businesses operating within Canada. The Members examine the potential benefits and drawbacks
of synthetic media in respect of their individual mandates (content production, competition,
privacy and copyright).
The purpose of this paper is to summarize to a general audience the perspectives of the four CDRF
Members on synthetic media. This report is not a policy position of the CDRF or its Members.
Instead, its main goals are to:
• inform Canadians of the variety of uses of gen AI and synthetic media
• provide an overview of the associated benefits and risks, as anticipated by CDRF Members
• provide key considerations for the technology as it develops into the future
Opportunities and Risks
As the technology behind synthetic media becomes more sophisticated and more accessible, its
growing use has benefits and risks. Some potential opportunities for Canadians include:
• Enhanced creativity and innovation: Gen AI systems have democratized the production of
creative outputs by allowing users the ability to generate synthetic content.
• Economic benefits: Synthetic media can reduce production costs and time, which may allow
businesses to reallocate resources to other areas and promote eiciency.
5
• Improvements to competition: In certain industries, synthetic media may enhance the ability
of smaller companies to compete with larger firms.
• Accessibility: Synthetic media can make content and services more accessible to people in a
variety of ways.
The widespread use of synthetic media and its increase in sophistication could also present
significant risks, including:
• Deceptive practices: Synthetic media could be used to mislead individuals by producing
convincing misinformation and disinformation for purposes such as marketing or journalism.
• Copyright infringement: The use of copyrighted material in the creation of synthetic media
may create legal uncertainty as the lines between the requirements to seek permission, fair
dealing and copyright infringement still need to be drawn, in Canada and internationally.
• Economic and employment consequences: Synthetic media and gen AI systems more
generally could lead to job displacement in certain sectors of the economy or a shift in revenue
from the artistic or creative sectors to the technology sector.
• Impact on reputation and privacy: Synthetic media may contravene privacy rights when
personal information is used as input in gen AI systems or when the generated output includes
information about an individual, risking negatively aecting their reputation, quality of life or
ability to be in control of their own personal information.
Introduction
Canadians are faced with a digital world accelerating at a breakneck pace. This is certainly true with
the rise of synthetic media. Canadians often see examples of the use of synthetic media whether it
is using AI voice-generating technology to create a song featuring two popular Canadian artists or
generating a fake video of the former Prime Minister appearing to promote a financial
scam. Worryingly, Canadians recently saw this technology used domestically to create a notorious
website for non-consensual, AI-generated pornography of real people.
The production and circulation of synthetic media has been the subject of many debates, ranging
from ethical and privacy concerns and potential misuse to its transformative possibilities in fields
such as culture, communication, education, and business.
What is synthetic media?
Synthetic media refers to artificially generated images, video, text, or audio content. Typically,
synthetic media refers to content produced using AI technologies. However, the term is also used in
cases where media has been generated using simple automation. This report is about the
implications of synthetic media generated using AI.
6
Gen AI has made the production and distribution of synthetic media possible on a larger scale.
Open AI’s ChatGPT and Meta’s Llama, also referred to as large language models or LLMs1, are
examples of gen AI models that have made the technology mainstream. In Canada, the popularity
of gen AI is rising: an October 2024 MTM 18+ analysis revealed that around 22% of Canadian
Internet users had used gen AI tools in the last month.
There are many dierent kinds of synthetic media, one of the most prominent examples being
deepfakes. Deepfakes are images, audio and videos that are digitally altered or generated by AI, to
depict real or fictional people or things.
International regulation and initiatives
In 2024, the World Economic Forum published its Global Risks Report presenting AI-generated
misinformation and disinformation as the second most likely risk to create a global crisis, behind
weather-related extreme events. However, governments and private businesses are continuing to
invest in AI, particularly in light of the technology’s potential benefits both for the economy and the
population’s quality of life. To mitigate potentially negative impacts and harness the benefits,
significant eorts to regulate AI are underway in various jurisdictions around the world.
One of the most notable eorts is the European Union’s Artificial Intelligence Act. The Act regulates
AI by requiring the analysis and classification of AI systems according to risk level and establishes
dierent requirements and obligations depending on that AI risk qualification. The Act prohibits
specific AI systems which have been determined to present risks that are considered unacceptable
(such as social scoring systems or manipulative AI). Other AI systems are designated as posing
either high, limited, or minimal risk.
High-risk AI systems (such as those used to profile individuals in areas such as education,
employment and law enforcement) are subject to stricter and more extensive obligations under the
Act, including requirements to develop risk management systems, ensure human oversight, and
implement explicit practices related to data governance.
While gen AI is not classified as high-risk as a starting point, it has to comply with transparency
requirements and copyright law, including disclosing that the content was generated by AI and
designing the model to prevent it from generating illegal content. The Act also clarifies that AI
developers must secure permission to use copyright protected work in developing and training AI
and gen AI systems as well as publish summaries of copyrighted data used for training.
Regulators and lawmakers from the United States and the United Kingdom have released guidance
and regulations specific to synthetic media. In the United States, regulators at the state-level have
developed legislation to curb the possible misuse of synthetic media for purposes like fraud,
election manipulation and non-consensual explicit content. In September 2024, the Government of
California passed a law specific to the media industry. The law protects the digital likenesses of
1 The Canadian Centre for Cyber Security defines LLMs as artificial neural networks that are trained on very
large sets of language data and enable users to complete sentences or generate entire documents by
receiving prompts.
7
performers and requires the informed consent of workers over the use of their voice, image or
likeness.
In the United Kingdom, the communications regulator, Ofcom, published a note on synthetic media
for the broadcasting sector outlining the challenges that might arise through the use of gen AI and
synthetic content. As well, the UK Digital Regulation Cooperation Forum published a paper, The
Future of Synthetic Media: A Summary of Stakeholder Views on the Future Development of
Synthetic Media and its Implications for DRCF Regulators, in November 2024.
Canadian Context and Environment
The Government of Canada has established initiatives around AI in recent years, with several
departments and agencies adopting guidelines and regulations requiring the responsible
development of AI. Examples include the Treasury Board Secretariat of Canada’s Directive on
Automated Decision-Making, the Government of Canada’s Guide on the use of generative artificial
intelligence, and the AI Strategy for the Federal Public Service 2025-2027.
The 2023 Voluntary Code of Conduct on the Responsible Development and Management of
Advanced Generative AI Systems has also provided businesses with a set of guidelines that they
can implement immediately. This code encourages businesses developing or managing gen AI
systems to achieve a set of positive outcomes that include accountability, safety, fairness, equity,
and transparency.
In the 2022-2023 period, Canadian AI research and development attracted over CAD$2.57 billion of
investments. Combined with a strong and competitive environment, these investments have
contributed to Canada’s being ranked fourth in per capita gen AI-focused companies operating in
the world.
In February 2025, Canada signed the Statement on Inclusive and Sustainable AI from the AI Action
Summit held in Paris. The Summit was an opportunity for jurisdictions to reairm their commitment
to encouraging innovation and deployment of open, inclusive, transparent, ethical, safe, secure and
trustworthy AI technologies.
Presentation of CDRF members’ sections
The Copyright Board’s section explores the questions surrounding the protection of copyright in a
world where the creation and dissemination of gen AI content is increasing exponentially. Issues of
importance are the attribution of copyright and remuneration of creators in synthetic media,
especially in the context of data-mining by LLMs. The legal framework varies from jurisdiction to
jurisdiction, and will most likely be clarified by the courts ahead of legislative changes, as live
issues are arising now between users and copyright owners.
The Board renders decisions regarding matters of which it is seized, which usually pertains to a
specific business or activity using copyright-protected works. As such, issues related to synthetic
media could make their way into the Board’s activities, impacting its tari setting proceedings and
on the licences granted when authors are unlocatable.
8
The CRTC’s section of this report explores the growing influence of gen AI and synthetic media on
Canada's communications and broadcasting sectors. It discusses the potential benefits and risks
of these technologies and their potential impact on Canadians, including creative workers and
journalists. It also examines considerations relevant to the CRTC’s mandate and activities, such as
the impact of synthetic media on the definition of Canadian content and on the provision of
accessible services.
The Competition Bureau’s section, while recognizing the beneficial application of synthetic media,
discusses how they can also be used to make deceptive marketing conduct more convincing. Using
hypothetical examples, the section provides a brief illustration of how synthetic media, particularly
deepfakes, can be misused to mislead and deceive the public. These misuses could raise concerns
under the deceptive marketing practices and Canada’s Anti-Spam Legislation (CASL) provisions of
the Competition Act.
An overview of three research-based considerations for synthetic media labels is also
included. These considerations are a potential strategy to help distinguish between content that is
synthetic and non-synthetic, thus empowering the public to make well-informed decisions.
The OPC’s section focuses on two main ways that privacy considerations are relevant to synthetic
media (i) as informational inputs used to both refine and hone algorithmic decision-making; and (ii)
as outputs of a synthetic media system, where privacy rights may be engaged.
The section covers relevant questions to ask when determining whether federal privacy law applies
to synthetic media systems. Finally, it presents relevant considerations when resolving how to
protect privacy rights in the creation and use of synthetic media systems.
Key Terms – Definitions from the Canadian Centre for Cyber Security (unless otherwise noted)
Natural language processing is a method to translate between computer and human languages. It
is a method of getting a computer to understandably read a line of text without the computer being
fed some sort of clue or calculation. In other words, natural language processing automates the
translation process between computers and humans.
Large language models (LLMs) are artificial neural networks that are trained on very large sets of
language data using self- and semi-supervised learning. LLMs initially generated text via next word
prediction but can now take prompts that enable users to complete sentences or generate entire
documents on a given topic. Training on exceptionally large datasets allows the model to learn
sophisticated linguistic structure, but also the biases or inaccuracies found in that data.
Machine learning (ML) is a field of research into methods that allow machines to learn how to
complete a task from given data without explicitly programing a step-by-step solution. ML models
can often approach or exceed human performance for certain tasks. As such, machine learning is
considered a sub-discipline of AI research.
9
Deepfakes are images, audio and videos that are digitally altered or generated by AI, to depict real
or fictional people or things. They represent a subset of the general category of “synthetic media” or
“synthetic content”, which includes AI-generated text.2
Voice cloning is a process that uses data sets from a single voice to train AI systems to copy the
speaking style, gender, age and accent of a specific individual’s voice. These AI systems can then
output new dialogue that sounds exactly like the original speaker.3
Social scoring systems are AI systems that evaluate or classify natural persons or groups thereof
on the basis of multiple data points related to their social behaviour in multiple contexts or known,
inferred or predicted personal or personality characteristics over certain periods of time. The social
score obtained from such systems may lead to the detrimental or unfavourable treatment.4
Synthetic media and the Copyright Board
The Copyright Act protects the intellectual property of creators, ensuring that they can control and
benefit from their work. In turn, this furthers the public interest by encouraging innovation and
creativity, allowing rights holders to be remunerated for their creations while ensuring access to
these works. These can include literature, music, art, films, software, and other forms of creative
expression.
2 For reference, certain other similar definitions exist, including Article 3 of the European Union’s AI Act,
which defines a deepfake as “an AI-generated or manipulated image, audio or video content that resembles
existing persons, objects, places, entities or events and would falsely appear to a person to be authentic or
truthful” (EU AI Act as of 13 June 2024); and the UK’s OfCom defines deepfakes as, “forms of audio-visual
content that have been generated or manipulated using AI, that misrepresent someone or something” (UK
OfCom (July 23, 2024), Deepfake Defences – Mitigating the Harms of Deceptive Deepfakes).
3 Papercup, ‘What is Voice Cloning?’ https://www.papercup.com/m/what-is-voice-cloning.
4 Definition adapted from Recital 31 of the EU AI Act.
10
The production of creative works by machines is not new.5 However, advancements in AI and
particularly gen AI, have evolved in part because of the availability of large amounts of machine-
readable digital content, including potentially large amounts of copyright-protected content.
Synthetic media is expected to radically alter the entire creative landscape, accelerating the
creation and delivery of content, significantly increasing the volume of content available.
For the purpose of this section, “synthetic media” refers to any content created, either fully or
partially, through production, manipulation, and modification of data or media, by automated
means (including the use of AI). This can include text, image, video, and sound content (“works” or
“copyrighted works”).
Use of copyright-protected content without permission is bound to take new forms as AI-based
business models emerge. Systems for tracking and recording the use of copyright-protected
content in synthetic media are being explored, but are not likely to be standardized, at least for
some time, and especially at the global level. Legal uncertainty in using data and information to
“feed” generative AI increases the risks related to these emerging business models and practices,
and responses to these risks will dier for each firm.
At this time, these legal and policy questions have only begun being tested or clarified. In the face
of a marketplace seeking to keep pace with emerging technologies, market uncertainty and uneven
domestic and global regulation may hurt both creators and users of copyright-protected content.
As an administrative tribunal and economic regulator, the Copyright Board is not responsible for
legislative or policy direction related to copyright. These functions are within the jurisdiction of the
Ministers of Industry and of Canadian Identity and Culture. That said, we do expect synthetic media
issues to surface in the Board’s activities very soon, in the same way that the first taris related to
Internet activities were filed with the Board in the mid-1990s. They were filed far ahead of the
widespread shifts in digital media production and consumption, and of any legislative changes that
are now fixtures of the current landscape.
As per the Copyright Act, the mandate of the Copyright Board is to approve taris for the use of
creative content protected by copyright when rights are managed by collective societies
(“collectives”). It settles individual cases where parties disagree and it grants licences for the use of
content when rights owners are unlocatable.
These functions are instrumental in facilitating marketplace activity, moderating the balance of
power in the marketplace, and protecting the public interest. The Board’s decisions aect a wide
range of businesses and industries, and touch sectors that impact the everyday life of all
Canadians, such as online music streaming, libraries, restaurants, fitness classes, airplanes, and
hotels.
Board decisions only apply to activities that take place in Canada; however, royalties generated by
taris are collected by Canadian collectives and distributed to their counterparts in other
countries, and vice versa. A number of global companies including Pandora Music, Apple and
5 Consider for example artworks developed by digital artists in the 1960s and 1970s using computer
technology.
11
Amazon participate regularly in Board proceedings. The following illustrates some of the ways
synthetic media could impact the Board’s activities.
Considerations
Impacts on Board proceedings
The Board’s decisions are typically made in reference to a specific business or activity in which
copyright-protected works are used. The Board anticipates the introduction of synthetic media
issues into its proceedings possibly next year, or the year after. There are a number of ways in which
AI or synthetic media may influence Board proceedings, such as:
• New taris: One or more new proposed taris could be filed by collectives, seeking
payment for the use of copyrighted content in training AI algorithms. For example, were
copies made of the works used to train the AI? Were these stored, and if so, for how long? In
such a case, the issues would be examined through the usual Board proceeding process
and would give an opportunity for both users and creators to argue their case.
• Repertoire considerations: Copyright collectives manage the rights of their members
regarding a specific protected right. Repertoire-use issues are often raised by users seeking
to confirm whom the collective represents, and how much of its repertoire is used in their
business activities. Questions may arise related to the repertoire of content covered by the
tari activities such as the inclusion of AI content, which might trigger additional
considerations like the identity of the author (human/machine or machine only) and if the
works are protected by copyright or might be infringing on other works.
• Unlocatable Owners: It is unclear at this time whether generative AI or synthetic media
would aect the requests received by the Board for licences when the owner of the
copyright is unlocatable. It is possible that the number of requests could increase,
particularly given an increased volume of content available overall. In any case, issues such
as authorship would also impact the consideration of the requests.
In all of these examples, it is expected that Board consideration of proposed taris and requests for
licences would entail tackling novel and highly contested issues through multi-party proceedings
and requiring ample evidence. It is also likely that evidence may be diicult to obtain, depending on
reporting and transparency requirements and practices – will data regarding AI creation and use be
tracked and reported, and if so, how? Current challenges in identifying data used to train AI’s
algorithms may be transposed to Board’s proceedings.
Legal and economic uncertainty
As an administrative tribunal and an economic regulator, the Board plays a role in clarifying the
legal framework of the creative marketplace as it interprets copyright legislation in a business
context.
As a tribunal of first instance, the Board is regularly called upon to apply the existing legal
framework to new activities and emerging technologies. This was the case with early Internet
12
activity, when taris for the distribution and reproduction of copyrighted works online were first
filed. When this happens ahead of policy direction or legislative changes, the Board’s decisions can
influence the legal and policy landscape. They have also prompted the pursuit of legal certainty
through appeals to the Federal Court of Appeal and the Supreme Court of Canada.
The legal uncertainty pertaining to synthetic media is increasingly a challenge to copyright owners’
businesses, especially in a commercial context. With this in mind, the Government of Canada held
a consultation on copyright and generative artificial intelligence, between 2023 and 2024. Key
issues identified in this consultation included:
• Whether and how copyright-protected content is legitimately used in generative AI training
• Whether and to what extent synthetic media is protected by local and international
copyright and can be monetized as such
• Whether and how the existing infringement and liability framework fit the AI environment
Along with legal uncertainty, practical ambiguities also generate growing concern in the Canadian
and global marketplace with respect to transparency in tracking the creation, use, and licensing of
content in an AI context. During the consultation, stakeholders raised a number of issues,
including:
• Text and data mining (TDM): TDM is an essential step in training AI systems that
reproduces and analyzes large quantities of data and information to identify patterns and
make predictions. Whether TDM activities require permission of copyright owners has not
yet been clarified in Canadian jurisprudence. Users are seeking an exception in the law that
supports the development of AI businesses. Many stakeholders were also interested in
transparency requirements (that is, record-keeping and disclosure requirements) to track
the use of copyrighted content by generative AI.
• Authorship of synthetic media content: As stated earlier, existing jurisprudence in
Canada suggests that authorship and therefore copyright protection can only be attributed
to a human who exercises skill and judgment to create a work. Overall, the consultations
demonstrated opposition to extending copyright protection to synthetic media without
suicient human contribution, but there was support for creators to use AI tools in the
creation of works.
• Infringement and liability: At this time, we are aware that at least one Canadian Court has
been seized to examine infringement with respect to AI-generated content. It is unclear
whether existing legal tests and remedies will be suicient. It is also unclear who would be
liable: developers, owners, or users of AI systems?
All these issues are relevant for Copyright Board proceedings, and any related court findings would
be likely to be taken into consideration in future Board proceedings.
13
International issues in copyright and synthetic media
As with all digital issues, synthetic media, copyright and AI have an inherently international reach.
Various jurisdictions and international bodies are developing their own views about these issues,
and what actions (if any) are warranted. Some, including Canada, have adopted AI voluntary codes
of conduct or ethical frameworks that contain principles that could aect copyright-based
industries. Discrepancies between national legislations on the global stage risk aecting the
protection and treatment of non-Canadian copyrights in Canada (and of Canadian copyright in
other countries), and the development of business models in the Canadian and international
marketplaces.
• The EU was the first to pass legal requirements with respect to AI, with the adoption of the
EU Artificial Intelligence Act. With respect to copyright, this Act clarifies that AI developers
must secure permission to use copyright protected work in developing and training AI and
generative AI systems.
• A bill aiming to assist rights holders in identifying works being used to train AI was
introduced to the US Congress in late 2024 (S.5379 TRAIN Act), signalling a shift to a
potential new legal framework. It should be noted that the future of this bill is uncertain
following the 2024 US elections.
We can expect that legal developments in the EU and the US will impact practices worldwide. It is
also expected that the legal framework around synthetic media will similarly be shaped by judicial
decisions, as parties bring their issues to the courts instead of waiting for legislative changes.
Just as copyright AI litigation is becoming more prevalent, a first Canadian case has been brought to
the B.C. Supreme Court by Canadian Legal Information Institute (CanLII) over the use of bulk
information by an AI chatbot. This could be the first case to bring light on the question of “AI
scraping” as copyright infringement in Canada, which could evolve in parallel with the litigation
from the New York Times against OpenAI (ChatGPT).6
It is unclear at this time how current international copyright treaties and existing trade agreements
to which Canada is signatory interact with AI and synthetic media. At the World Intellectual
Property Organization (WIPO), proposals to study copyright in the digital environment have been
put forward since 2015 and have led to a number of studies and reports.
In 2024, several proposals were made to study more specifically AI and regulatory challenges,
market practices and to organize information sessions on these issues. That said, amending
international treaties or creating new ones sometimes takes even longer than national legislations.
The evolving technological landscape is likely to bring forth novel issues. One example is the
authorship of works. The Berne Convention, of which Canada and 180 other countries around the
6 The New York Times Company v. Microsoft Corporation, 1:23-cv-11195 (S.D.N.Y. - 2023).
https://www.courtlistener.com/docket/68117049/the-new-york-times-company-v-microsoft-corporation/.
14
world are signatories, specifies that copyright protection is to be attributed to the author upon
creation, without any requirement for registration.7
As this question takes on an entirely new dimension with synthetic media, various countries may
start to reconsider the costs and benefits of registration. To remain manageable, these issues
would require international agreement, or risk devolving into a complex and unmanageable system
where protection varies across countries. Such a scenario would be detrimental to both creators
and users of copyright-protected content.
Final observations
The creation and dissemination of content generated by AI poses many challenges from the point of
view of copyright. Synthetic media has already permeated the way that content is made and shared
online and as is often the case with many revolutionary technologies, the policy and legal
frameworks have yet to catch up with the challenges posed.
The Copyright Board finds itself at an interesting point in the copyright ecosystem, as its role is both
one of legal adjudicator and economic regulator. As a tribunal of first instance, it is at the forefront
of legal developments applicable to the use of copyrighted works when novel issues aect
collective societies (representing copyright owners) and users. While its power is contingent on the
Board being seized of matters and on applying the existing legal framework, it is expected that the
Board will play a role in this new era of creation, reproduction, and distribution sooner rather than
later.
Synthetic media and the Canadian Radio-television and
Telecommunications Commission
In the last few years, worldwide use of gen AI and synthetic media has surged across several
industries, including in various areas of the communications sector. The CRTC is an independent
quasi-judicial tribunal that regulates the Canadian communications sector. Although the CRTC
does not regulate AI or gen AI systems, their possible use within the communications sector could
affect the services and industries that it regulates. As such, the CRTC has had to stay informed of
developments related to gen AI and synthetic media. It is vital that the use of gen AI and synthetic
media within the communications sector respects and upholds Canadian values.
This section considers the impacts of gen AI and synthetic media on the broadcasting industry as
well as on unsolicited communications. It provides an overview of the key issues and recent
developments, as well as the potential benefits and risks of gen AI and synthetic media observed to
date by the CRTC.
7 “The enjoyment and the exercise of these rights shall not be subject to any formality” Berne Convention for
the Protection of Literary and Artistic Work, art. 5, para.2, September 28, 1979.
15
Considerations
Impacts on the broadcasting industry
The CRTC is modernizing Canada’s broadcasting framework to help ensure that Canadians have
access to Canadian and Indigenous content through a variety of platforms. In this context, the key
issues and questions for the CRTC regarding gen AI and synthetic media are:
• Impact on the audio-visual and audio broadcasting sectors;
• Impact on the creation of Canadian Content; and,
• Impact on the creation and availability of news.
As a quasi-judicial administrative tribunal, the CRTC makes decisions based on the public record.
The CRTC’s public consultations and hearings allow it to hear from a diverse set of voices. These
voices include: individuals; representatives of First Nations, Inuit, and Métis; public interest groups;
and industry. The CRTC is gathering public input on gen AI and synthetic media across its various
consultations and hearings as it seeks to help modernize Canada’s broadcasting framework,
including in:
• Broadcasting Notice of Consultation CRTC 2024-288 - The Path Forward – Defining
“Canadian program” and supporting the creation and distribution of Canadian programming
in the audio-visual sector (NoC 2024-288); and
• Broadcasting Notice of Consultation CRTC 2025-52 - The Path Forward – Supporting
Canadian and Indigenous audio content (NoC 2025-52).
Across these consultations, the CRTC is asking exploratory questions on whether gen AI materials
could be considered Canadian content as well as on the potential impact of gen AI on content
production and on the discoverability and promotion of Canadian content.
Impact on the audio-visual and audio broadcasting sectors
Canadian creators – such as actors, musicians, writers, and producers - create a wide range of
programming that reflects Canadian attitudes, opinions, ideas, values, and artistic creativity. The
stories they create reflect the needs and interests of Canadians and provide information to
Canadians on important current issues. They also reflect diverse voices from Indigenous, equity-
seeking and French-language communities, enriching the Canadian broadcasting system and
wider economy.
With the advent of gen AI and synthetic media, Canadian creators have new tools at their disposal
to generate audio-visual and audio content. For example, gen AI and synthetic media can be used
to create special effects, complete parts of a movie script, design a lifelike movie poster in
seconds, or produce digital replicas.
Some stakeholders in the audio-visual and audio broadcasting sectors are welcoming gen AI and
synthetic media, since the benefits are well known to creative workers and entertainment
companies alike. In its 2023 submission to Parliament, Alliance of Canadian Cinema Television and
16
Radio Artists (ACTRA) recognized the potential for gen AI to complement the work of creators and to
reduce the costs associated with casting, dubbing and subtitling, digital transformation, and
motion capture.
For example, the 2024 Coca-Cola holiday commercial was created by using five different gen AI
services to produce 110 versions of the commercial within a few days and at a lower cost.
Gen AI and synthetic media might help businesses become more efficient or comply with
regulations. For example, as part of its implementation of the amended Broadcasting Act, the CRTC
held a public proceeding to learn more about how Canadians who rely on closed captioning can
access barrier-free programming on streaming platforms.
On the record of the proceeding, certain media companies reported the use of automated speech
recognition to allow for the captioning of more content at a lower cost. Similarly, the use of gen AI
and synthetic media in the dubbing industry is allowing filmmakers to re-dub content in post-
production.
This could have implications in Canada, because of linguistic diversity. For example, in the co-
development process of the Indigenous Broadcasting Policy, Indigenous peoples of Canada shared
with the CRTC that more radio and television content could be made available in Indigenous
languages through dubbing or subtitling. However, language translation can be expensive, and, in
some cases, speakers with native proficiency may be rare or not easily accessible. Gen AI and
synthetic media have the potential to bridge existing gaps for Indigenous communities.
At the same time, the adoption of gen AI and synthetic media in captioning and dubbing could lead
to potential challenges in employment. Job displacement in the creative industries is of concern
not only for the economy, but also for the broadcasting ecosystem as Canadian creators contribute
to the cultural vibrancy and diversity of Canada. Some organizations have highlighted that these
technologies threaten the livelihoods of the 750 artists in Quebec who contributed to 501 film and
television productions dubbed in French in 2023-2024. Of the 8,450 active members of Quebec’s
Union des Artistes, 44% also provide services in narration, video description, and voice acting. All
of these sectors could be impacted by these technologies.
Canadian screenwriters took steps to protect their rights against the use of copyrighted materials to
train AI systems without their expressed consent, encouraged by the protections American
creatives gained from the Writers Guild of America’s strike. In May 2024, the Writer’s Guild of
Canada (WGC) concluded a three-year agreement with the Canadian Media Producers Association
(CMPA) that requires producers to disclose the use of AI-generated materials when providing them
to writers. It also prohibits a reduction in existing writer credits or compensation.
Some professional guilds have indicated that gen AI poses a generational threat to their members’
livelihoods and that it has already impacted the cultural industries. Performers also want to
consent, earn compensation, and control how AI models use their name, image, and likeness. The
most recent Independent Production Agreement between ACTRA, the CMPA and the Association
Québécoise de la Production Médiatique (AQPM) now includes rules protecting performers from AI-
generated material, digital replicas and digital alterations used or made without their expressed
consent.
17
Impact on the creation of Canadian Content
The Online Streaming Act received Royal Assent in April 2023 and became the first major revision to
the Broadcasting Act since 1991. The CRTC is responsible for the implementation of the
modernized Broadcasting Act and has, in its first 24 months of implementation, held 13 public
consultations and issued four decisions. Streaming services operating in Canada are now expressly
included in the country’s broadcasting framework to ensure they contribute to the creation and
distribution of Canadian content, among other things.
In November 2024, the CRTC issued NoC 2024-288 to modernize the definition of Canadian
programming in the audio-visual ecosystem8 and invited comments on:
• whether AI-generated material can be considered “Canadian content”
• the possible impacts of AI on pre- and post-production
• how AI could impact the discoverability of Canadian content
The CRTC is not the only organization responsible for certifying content as Canadian. The Canadian
Audio-Visual Certification Office (CAVCO), Telefilm Canada and the Canada Media Fund can also
certify content for funding and tax credits, although each organization uses different criteria.
The CRTC certification of Canadian film and television programs is important because it is linked to
the funding of Canadian programs through mechanisms such as Canadian programming
expenditures9 and certified independent production funds.
The potential for programs that make use of gen AI and synthetic media to qualify as "Canadian
content" and, in turn, receive public funding, is being considered by some professional guilds. For
instance, the WGC included protections in their production contracts with the CMPA to ensure that
AI-generated content provided by producers to writers or story editors would not receive
compensation or credit.
Gen AI and synthetic media also have the potential to impact Canadian content requirements for
radio and Canadian content requirements for television. Campus radio stations, for example, must
comply with specific requirements to maintain their broadcasting licences, including an obligation
to devote no less than 15% of their programming to locally produced news or spoken word content.
In a 2023 press release, the National Campus and Community Radio Association (NCRA)
acknowledged that gen AI and synthetic voices could be used by smaller stations with limited
resources to create hourly local news and weather updates. The NCRA argued that “the use of AI
content in spoken word programming is where we predict it will be noticed first in our sector.”
While there are no regulations on the use of synthetic media in radio programming as the
technology is still recent and developing, the NCRA advised its members of the possible biases of
8 The current definition of a Canadian program is based on specific criteria and a point system, which focus
on the financial and creative control of productions.
9 Canadian programming expenditures are used to contribute to the representation of a diversity of voices in
the Canadian national broadcasting system. Canadian broadcasters are required to allocate portions of their
annual broadcasting revenues to expenditures on Canadian programming.
18
AI and its capacity to produce false or misleading news. The uses and potential impacts of AI in the
audio sector are currently being studied by the CRTC in NoC 2025-52 to modernize its policy for
radio and audio streaming in Canada.
Impact on the creation and availability of news
The CRTC does not play a direct role with respect to what news content is produced or with
journalism more generally. Instead, its role in the news sector is primarily to ensure that Canadians
have access to high-quality news and current affairs programming. News services are considered
programming of exceptional importance and the CRTC can require mandatory distribution of news
programming. The CRTC also ensures that certain broadcasters make and broadcast news through
various incentives and that others contribute to supporting the cost of news production. In
addition, the CRTC has a role in implementing parts of the Online News Act, which aims to ensure
that online platforms that make Canadian news content available fairly compensate Canadian
news organizations.
While not directly impacting the mandate of the CRTC, AI is adding a further dimension to the
already complex relationship between news organizations and online platforms. For example, a
coalition of Canadian news publishers recently filed a lawsuit against OpenAI claiming that it uses
copyrighted material from Canadian news media while preventing journalists from earning
important revenue at a time when Canada’s news industry is experiencing financial difficulties.
Canadian newsrooms have also begun using gen AI and synthetic media to reduce operational
costs. An October 2024 KPMG survey on the use of AI of more than 800 Canadian news
organizations showed that 61% have implemented gen AI tools and 89% describe the technology as
“extremely or very important to their competitive advantage.”
Journalists, writers, and editors are using gen AI and synthetic media to produce stories faster or to
focus on investigative journalism while gen AI accomplishes routine tasks (for example, transcribing
and summarizing city council meetings).
However, gen AI tools can also be used to produce fake news stories that appear legitimate. These
tools can quickly generate highly convincing synthetic media that could deceive Canadians. The
Global Journalism Innovation Lab surveyed 1,042 Canadian respondents and found that 70% were
concerned about inaccuracies in news stories due to AI, while 78% expressed worries about gen
AI's potential to spread misinformation and disinformation.
Tools that limit the spread of misinformation or disinformation, such as synthetic media, are
important to Canadians. In a 2023 study conducted by the Canadian Journalism Foundation, half of
the Canadians surveyed indicated that they are not confident in their ability to distinguish between
AI-generated content and human-created content. Another 85% of respondents in the survey
conducted by the Global Journalism Innovation Lab expressed a desire for newsrooms to be
transparent about their use of AI, with three quarters advocating for the labelling of AI-generated
content.
19
Final observations
The CRTC’s section of this report discussed the intersection between its mandate, the use of
synthetic media, and some of the potential positive and negative aspects of AI in the audio-visual
and audio broadcasting sectors. AI may help creators and reduce operational costs. However, AI
may also have negative impacts.
Further monitoring will be needed to understand the potential drawbacks of gen AI and synthetic
media, such as the use of AI in ways that are not transparent to the public.
Digital regulators around the world can play an important role in shaping the effective and
responsible use of gen AI and synthetic media. For example, digital regulators can collaborate and
consult with the public, industry stakeholders and other governmental organisations to share
knowledge and align principles, such as those related to accountability, transparency, and
accessibility.
The fast-paced nature of AI advancement may also point to a need for flexible regulatory
approaches. In some jurisdictions, such as the European Union, industry stakeholders have been
calling for the development of guidelines and principles to frame the use of gen AI in the creative
industries.
Given the global reach of AI technologies, regulatory agencies’ cooperation through knowledge
sharing and other forms of collaboration will remain important to the development of proactive and
flexible digital policies. This cooperation can take place domestically, through forums such as the
CDRF, and internationally, where regulators can benefit from the exchange of information, learn
about initiatives, and share best practices. Although they are still emerging, the AI technologies
responsible for the production and diffusion of synthetic media are poised to present the Canadian
telecommunication and broadcasting industries, and consequently, the CRTC, with challenges and
opportunities alike. It is vital that the use of gen AI and synthetic media within the communications
industries respects and upholds Canadian values, and the CRTC will have a role in ensuring this.
Synthetic Media and the Competition Bureau
This section provides an overview of how the use of deepfakes can violate the Competition Act’s
deceptive marketing practice provisions.
The Competition Act states that it is against the law for businesses and individuals to advertise or
market goods, services or any business interests to the public in a way that is materially false or
misleading. This is because such practices can harm consumers, businesses and competition, and
negatively impact the economy.
The provisions apply to all promotions made through:
• printed or electronic advertisements
• written or oral representations
• illustrations
20
• audio-visual promotions
They also apply regardless of the medium used to make the promotion.
To ensure that consumers can make informed decisions, truthful and accurate information is
necessary in all advertising media.
For illustrative purposes, included below are descriptions of certain deceptive marketing practice
provisions and examples of deepfake usages that may raise an issue under each of these
provisions. Note that none of the deepfake examples have been found to violate the Competition
Act by a court or tribunal.
Considerations
Deepfakes and the Competition Act’s deceptive marketing
provisions
Competition Act
Provision
Provision Description and Examples
Section 52 [criminal
provision], and
Paragraph
74.01(1)(a)
[civil
provision]
False or misleading representations
It is against the law to make
materially false or misleading
representations to promote a product, service or any business interest.
AI can exacerbate some existing deceptive marketing practices by
enabling bad actors. For instance, the technology may be used to make
false or misleading representations more convincing with minimal eort,
or to achieve a much greater scale with very lit
tle cost. See the
Competition Bureau’s page on
common scams and deceptive marketing
practices and how to avoid them
.
Example
: Deepfakes can make fake reviews and endorsements seem
authentic. For example, a business can use deepfakes of well
-known
celebrities or public figures without their permission to create marketing
videos with fake endorsements or fake reviews of a product
.
Paragraph
74.01(1)(b)
[civil
provision]
Performance claims that are not based on an adequate and proper
tests
A business who makes a claim about the performance, eectiveness or
length of a product or service’s life must be able to prove that the claim is
based on an adequate and proper test.
Potentially unsubstantiated performance claims of products and
services, including those relating to AI features, can cause harmful
eects. Outcomes may include honest competitors losing business and
21
consumers not having the information that they need to make informed
purchasing decisions.
Example
: Deepfakes of well-known and trusted individuals may be used
to make unsubstantiated claims more compelling. For example, an
unauthorized deepfake of a celebrity can be created to falsely promote
an investment scheme with an unsubstantiated performance claim, such
as “You can earn a 300% return within 30 days)
.”
Section 52.1
[Criminal provision]
Deceptive telemarketing
In telemarketing, it is a criminal oence to fail to make certain
disclosures at the beginning of each communication. It is also a
criminal
oence to make or allow the making of materially false or misleading
claims to promote a product, service, or any business interest when
communicating orally over the phone or by using any form of
telecommunication including recorded messages and
robocalls. A
robocall is a call made from an automated source with a prerecorded
message.
Example
: “Deepfaked” voices can make deceptive telemarketing more
eective and harder to detect. In deceptive telemarketing cases, bad
actors often falsely claim that they are calling from a certain company. AI
tools can allow them to create highly realistic voi
ce simulations, or
impersonate company executives, colleagues or trusted external
contacts during the calls.
They could also impersonate banks, government agencies and utility
firms through AI enabled robocalls, and request that consumers provide
personal information. They may also falsely claim to be calling about
purchase authorizations for orders allegedly pla
ced by the targeted
individuals.
Using behavioural science to understand the challenges and
limitations of synthetic media labels
The content below is not an expression of either the Competition Bureau’s endorsement of
or opposition to the use of labels as a strategy for addressing the potential risks of synthetic
media.
Background
It is critical to ensure that consumers are empowered to make well-informed decisions, a key
ingredient in a competitive marketplace. Therefore, it is a growing concern that consumers are
22
increasingly unable to distinguish between content that is synthetic and non-synthetic.10 While
academics, industry experts, and governments around the world are exploring several ways to
address the issue, one of the most discussed is labelling. Labelling is the addition of informative
tags to synthetic content (including deepfakes) that consumers can see or hear.11
Research has found many nuances to the eectiveness of labels. Some labels have been found to
be eective in reducing belief in false or misleading content and the sharing of such content
online.12 Additionally, some research indicates that labels have advantages over general awareness
campaigns. This is because these campaigns fail to improve consumers’ abilities to recognize
synthetic content and can even lead to skepticism about all media, including non-synthetic
media.13
However, many factors can influence a label’s eectiveness, and even well-designed labels may
sometimes result in only modest impacts relative to other types of interventions. See Labeling AI-
Generated Content: Promises, Perils, and Future Directions.
Furthermore, labels can even have negative, unintended consequences such as inappropriately
increasing trust in unlabelled content or failing to communicate the correct information as further
elaborated in the next section.
Considering these issues, it is important to weigh the pros and cons when thinking about
introducing synthetic media labels. Scientific research can shed light on labels’ challenges and
limitations as a strategy for addressing the potential misleading eects of synthetic media and
deepfakes. Below is a brief overview, prepared by the Competition Bureau’s Behavioural Insights
Unit, of three research-based considerations for synthetic media labels.
Three considerations for the use of labels
The following three considerations14 help illustrate some of the challenges of using synthetic media
labels and the possible limitations to their eectiveness that may be unavoidable.
10 Groh, M., Sankaranarayanan, A., Singh, N., Kim, D. Y., Lippman, A., & Picard, R. (2024). Human detection of
political speech deepfakes across transcripts, audio, and video. Nature Communications, 15(1), 7629.;
Köbis, N. C., Doležalová, B., & Soraperra, I. (2021). Fooled twice: People cannot detect deepfakes but think
they can. Iscience, 24(11).; Lewis, A., Vu, P., Duch, R. M., & Chowdhury, A. (2023). Deepfake detection with
and without content warnings. Royal Society Open Science, 10(11), 231214; Mai, K. T., Bray, S., Davies, T., &
Griin, L. D. (2023). Warning: Humans cannot reliably detect speech deepfakes. Plos One, 18(8), e0285333.
11 Bennet, M. (2023). Bennet Urges Digital Platforms and AI Developers to Label AI-Generated Content, Stop
the Spread of Misinformation.; Goujard, C. (2023, June 5). EU wants Google, Facebook to start labeling AI-
generated content. POLITICO.; Torres, R. (2023). U.S. Rep. Ritchie Torres Introduces Federal Legislation
Requiring Mandatory Disclaimer for Material Generated by Artificial Intelligence. Torres.
12 Martel, C., & Rand, D. G. (2023). Misinformation warning labels are widely eective: A review of warning
eects and their moderating features. Current Opinion in Psychology, 101710.; Martel, C., & Rand, D. G.
(2024). Fact-checker warning labels are eective even for those who distrust fact-checkers. Nature Human
Behaviour, 8 (10), 1957–1967.
13 Wittenberg, C., Epstein, Z., Berinsky, A. J., & Rand, D. G. (2024). Labeling AI-Generated Content: Promises,
Perils, and Future Directions.
14 Other challenges beyond the three considerations highlighted here remain. For example, the visibility and
timing of labels can also critically impact a label’s eicacy.
23
1. Consumer interpretations of label wording: Depending on a label’s wording and design, it
may or may not communicate what is intended. For example, a label could be intended to
communicate either a content’s truthfulness or its origin (synthetic or non-synthetic).
Behavioural science research has investigated how dierent label wording communicates
these messages about synthetic media.
One study found that consumers associate the label “AI Generated” with content generated
by AI more consistently than labels like “Synthetic” or “Deepfake”.15 At the same time,
consumers do not associate “AI Generated” with misleading content; therefore, the label “AI
Generated” prompts less skepticism than the labels “Synthetic” or “Deepfake”.16 Another
study found that AI-generated messages that are expressly labelled as “Made by AI” can be
just as persuasive as unlabelled messages, again showing how a label may communicate
content’s origin without prompting skepticism.17
Given these nuances to how consumers interpret and react to label wording, the available
research suggests that care should be taken when determining the purpose of any synthetic
media labels. Proper testing may help to ensure that the labels communicate their intended
purpose.
2. Consequences of labelling for unlabelled content: If the labelling of synthetic media
becomes more common, consumers might assume that unlabelled content is non-
synthetic.18 As a result, labelling synthetic media could have the unintended eect of
increasing consumer trust of unlabelled content, thereby potentially increasing the
victimization of consumers by unlabelled, false or misleading synthetic content.
However, research on misinformation has found that introducing labels for both true and
false content simultaneously can help to reduce an unintended increased trust in
unlabelled content.19 This evidence may support the position of some AI experts who
believe that, in this age of prolific synthetic media, the labelling of non-synthetic content as
verified may be helpful to consumers.20
3. Impact of repeated exposure: Seeing the same content repeatedly online may reduce the
eectiveness of any accompanying labels. As such, seeing false content repeatedly can
increase belief in its truth, even when the content is labelled with a warning. For example,
synthetic media that is labelled as such may nevertheless become more trusted or
15 Epstein, Z., Arechar, A. A., & Rand, D. (2023). What label should be applied to content produced by
generative AI?
16 Ibid.
17 Matz, S. C., Teeny, J. D., Vaid, S. S., Peters, H., Harari, G. M., & Cerf, M. (2024). The potential of generative AI
for personalized persuasion at scale. Scientific Reports, 14(1), 4692.
18 Pennycook, G., Bear, A., Collins, E. T., & Rand, D. G. (2020). The Implied Truth Eect: Attaching Warnings to
a Subset of Fake News Headlines Increases Perceived Accuracy of Headlines Without Warnings.
Management Science, 66(11), 4944–4957.
19 Ibid.
20 The Canadian Press. (2024, October 31). Fake content is getting harder to suss out. This Canadian Nobel
Prize winner has an idea to help. cbc.ca.
24
perceived as real after repeated exposures.
One study observed a related eect with false or misleading information where people who
repeatedly encountered the same false and misleading content came to have greater belief
in its truth, even when the content was labelled with a warning.21
Another challenge is that consumers often forget the source of labelled content.22 Existing
research suggests that as the proportion of media that is synthetic increases, consumers
may be more likely to falsely remember non-synthetic content as being synthetic.23
This research suggests that other strategies that limit the number of exposures to the same
content, could help labels maintain their eicacy. One example of such a strategy is
downranking (reducing the prominence content is given in a news feed or search results).
Final observations
Our review of existing research on labels suggests that labels may be a useful strategy for
addressing the potential risks of synthetic media. However, the eectiveness of labels depends
upon on the specifics of how they are implemented.24 In some cases, labels may introduce new
challenges or have unintended consequences. Finally, synthetic labels face other practical
challenges, such as non-compliance or fraudulent use to intentionally mislead or deceive.
In short, research suggests that labels are not a perfect solution and implementing them would
come with many challenges. Nevertheless, in some situations, they may be a tool in mitigating the
risk of synthetic media having misleading or anticompetitive eects.
21 Pennycook, G., Cannon, T. D., & Rand, D. G. (2018). Prior exposure increases perceived accuracy of fake
news. Journal of Experimental Psychology: General, 147(12), 1865.
22 Bell, R., Mieth, L., & Buchner, A. (2022). Coping with high advertising exposure: A source-monitoring
perspective. Cognitive Research: Principles and Implications, 7(1), 82.
23 Bell, R., Mieth, L., & Buchner, A. (2022). Coping with high advertising exposure: A source-monitoring
perspective. Cognitive Research: Principles and Implications, 7(1), 82.; Bell, R., Mieth, L., & Buchner, A.
(2020). Source attributions for detected new items: Persistent evidence for schematic guessing. Quarterly
Journal of Experimental Psychology, 73(9), 1407–1422. Kuhlmann, B. G., Vaterrodt, B., & Bayen, U. J. (2012).
Schema bias in source monitoring varies with encoding condi�ons: Support for a probability-matching account.
Journal of Experimental Psychology: Learning, Memory, and Cognion, 38(5), 1365.
24 Martel, C., & Rand, D. G. (2023). Misinformation warning labels are widely eective: A review of warning
eects and their moderating features. Current Opinion in Psychology, 101710.; Wittenberg, C., Epstein, Z.,
Berinsky, A. J., & Rand, D.G. (2024). Labeling AI-Generated Content: Promises, Perils, and Future Directions.
25
Synthetic media and the Oice of the Privacy
Commissioner
As the digital age comes into maturity, emerging technologies have raised important questions and
implications related to privacy. The phenomenon of synthetic media is no exception25; whether it be
related to deepfakes, voice cloning technology or the outputs of large language models (LLMs).
Privacy is of utmost importance given the amount of information that is needed, for example, to
train synthetic media algorithms to:
• generate accurate or convincing images;
• precisely mimic the timbre and inflection of a person’s voice based on a short reference
sample; or
• produce writing in the style of a well-known author or thinker.
The creation and use of this technology may impact privacy rights and prompts several important
privacy-related considerations.
How the federal privacy law for organizations applies to synthetic
media
From a privacy perspective, one of the first significant questions is how Canadian privacy law
applies to synthetic media. Canada’s private sector privacy law, the Personal Information
Protection and Electronic Documents Act (PIPEDA), applies to organizations across Canada that
collect, use, or disclose personal information in the course of a commercial activity. Notably,
Alberta, British Columbia and Quebec have their own private sector laws that have been deemed
substantially similar to PIPEDA. Organizations operating solely within one of these provinces would
therefore generally be subject to that province’s legislation instead of PIPEDA. PIPEDA only applies
to organizations in the private sector and does not apply to an individual acting for their own
personal, non-commercial purposes.
Many synthetic media organizations fall under PIPEDA if they collect, use and disclose personal
information in the course of commercial activities. See the OPC’s Interpretation Bulletin:
Commercial Activity. For instance, the creator of a synthetic media system selling to a third party, or
a third party inputting personal information into a synthetic media system to provide a service for
customers, would both likely be subject to PIPEDA.
This section examines synthetic media through the lens of PIPEDA and discusses both the
informational inputs and outputs of these systems. Since the application of PIPEDA is context
dependent, these examples are meant to be indicative, not determinative of any specific context or
set of facts.
25 Note: this term does not include synthetic data, which has its own definition and privacy implications.
26
Considerations
Inputs to Synthetic Media Systems
Definition of Personal Information
Synthetic media content is generally developed using very large data sets to refine algorithmic
systems to either create accurate imitations (that is, of an image or a voice), or to generate new
forms of content.
From a privacy perspective, this process raises several considerations. The first is whether the data
sets used to train synthetic media systems contain any personal information. While some may not
be trained using any personal information, certain kinds of synthetic media, such as deepfakes,
often use personal information to replicate images accurately.
In the context of deepfake production, information input into a synthetic media system with the
objective of accurately producing similar images in a dierent context would likely be considered
personal, as the image or images necessarily would need to be identifiable for the deepfake to be
similar to any extent. This would also likely apply to voice cloning, where data sets from an original
voice are used to train AI systems to copy the speaking style, gender, age and accent of a specific
individual’s voice.26 This is increasingly being raised as an issue when used by scammers and
cybercriminals to convince people to send them money or personal information.27
PIPEDA specifies that personal information is information ‘about’ an identifiable individual. This
includes things like pictures of an individual, their employment history, their driver’s license
number and even their DNA. In this sense, ‘about’ means that the information is not just the subject
of something but also relates to or concerns the subject. Furthermore, information concerns “an
identifiable individual” where there is a serious possibility that an individual could be identified
using that information, alone or in combination with other information.28
Related to voice cloning, in its guidance on personal information, the OPC has included
voiceprints29 as a form of biometric information that is considered personal, and in many cases can
even be sensitive information. See the OPC’s, Interpretation Bulletin: Personal Information. In this
sense, there are many scenarios where information input into a synthetic media system is
considered personal under privacy law.
Sensitive information
Whether the personal information input into a synthetic media system is sensitive is another
important consideration. While organizations may themselves input sensitive information into
26 Federal Trade Commission, (2024). Approaches to Address AI-enabled Voice Cloning. Federal Trade
Commission.
27 Aslett, D., Lin, L., Mekonnen, G.T., Zecevic, M., (2024), The dangers of voice cloning and how to combat it,
The Conversation.
28 Gordon v. Canada (Health), 2008 FC 258.
29 Wansink v. Telus Communications Inc., 2007 FCA 21, para. 3. “Voiceprints are … a matrix of numbers that
represent the characteristics of the employee’s voice and vocal tract. They are not considered the voice of the
person themselves.”
27
these systems, there may also be a risk that developers use very large datasets that may include
sensitive information to train their models. This could lead to risks to privacy.
Under PIPEDA, any personal information can be sensitive depending on the context. That said, the
OPC has found that certain types of personal information will generally be considered sensitive
because of the specific risks to individuals associated with the collection, use or disclosure of
these categories of information.
The OPC has outlined that information that “…will generally be considered sensitive and require a
higher degree of protection includes health and financial data, ethnic and racial origins, political
opinions, genetic and biometric data, an individual’s sex life or sexual orientation, and religious or
philosophical beliefs.” See the OPC’s, Interpretation bulletin: sensitive information.
Children’s information is also generally considered to be sensitive and is at higher risk of significant
negative impact. See the OPC’s Principles for responsible, trustworthy and privacy-protective
generative AI technologies.
The level of sensitivity impacts a number of requirements under PIPEDA, including whether or not
express consent is required for the collection, use or disclosure of the personal information, the
level of safeguards necessary to protect it, as well as whether a breach creates a real risk of
significant harm. As such, developers and operators must make sure that they know the level of
sensitivity of the information being input into a synthetic media system.
Consent
In most instances, PIPEDA will apply when an organization is collecting or using personal
information to train or use synthetic media-generating technologies. As such, organizations should
understand their Canadian privacy law obligations before designing and implementing synthetic
media systems, which generally include obtaining valid and meaningful consent when collecting
and using personal information.
Meaningful consent is an essential element of Canadian private sector privacy legislation. Under
PIPEDA, organizations are generally required to obtain meaningful consent for the collection, use
and disclosure of personal information. For consent to be meaningful, individuals must understand
what they are consenting to (that is, the nature, purpose and risks of harm or consequences
resulting from the personal information’s collection, use or disclosure). There may be limited
exceptions to consent or other legal grounds for collecting, using or disclosing this information,
depending on the circumstances.
The form of consent may vary and is impacted by the circumstances and the type of information. As
already noted, the sensitivity of the information being input into a system is a factor in determining
the form of consent30 that is required.31 Sensitive information generally requires express consent
from the individual to whom it relates. In a scenario in which sensitive personal information is
30 Personal Information Protection and Electronic Documents Act, S.C. 2000, c.5 (see Schedule 1, Principle
4.3.4 and 4.3.6).
31 Ibid (see Schedule 1, Principle 4.3.4).
28
inputted into a synthetic media system for the creation or generation of new images, videos, or text,
it may require the consent of the individual to whom it belongs or relates.
Publicly available personal information
A common misconception is that personal information that is available on the Internet may be
freely collected by organizations, including to either train, or input into, algorithms for synthetic
media systems. Just because personal information is publicly accessible online does not mean,
however, that it is ‘publicly available’ under Canadian privacy law.32 PIPEDA regulations provide for
a narrow definition of ‘publicly available information.’33
Since a data collector is responsible for determining whether consent is needed, they must know
whether the personal information that is inputted into a synthetic media system was scraped from
the web. As well, they must obtain appropriate and meaningful consent from individuals where
such consent is required.
Operators of websites that host personal information also have obligations under privacy law to
safeguard personal information on their platforms, to protect it from unlawful data scraping.
Moreover, data scraping incidents that harvest personal information may constitute reportable data
breaches for the holder of the data. See OPC’s Joint statement on data scraping and the protection
of privacy.
Accountability
Synthetic media raises important questions for organizations relating to accountability, which is a
central and fundamental PIPEDA principle that is connected to and underpins all other principles
and obligations.
PIPEDA states that an organization is responsible for personal information in its possession or
custody, including information that has been transferred to a third party for processing.34
Organizations need to implement policies and practices to, among other things, protect personal
information, receive and respond to complaints and inquiries, and develop information to explain
these policies and practices as part of their accountability responsibilities under PIPEDA.
Organizations should recognize their responsibility for compliance with privacy legislation and be
able to demonstrate compliance. The OPC recommends that organizations develop a privacy
management program to help correctly identify privacy-related obligations and risks and
appropriately take them into account before they launch new products or services. Tracing
32 For example, the OPC noted in its investigation of Clearview AI that ‘information collected from public
websites, such as social media or professional profiles, and then used for an unrelated purpose, does not fall
under the ‘publicly available’ exception’ of PIPEDA.
33 The Regulations Specifying Publicly Available Information (SOR/2001-7) made under PIPEDA specify that
publicly available information includes, amongst other things, (i) personal information that appears in a
telephone directory that is available to the public (ii) personal information of an individual that appears in a
professional or business directory, listing or notice, and (iii) personal information that appears in a
publication, including a magazine, book or newspaper, in printed or electronic form, that is available to the
public, where the individual has provided the information.
34 Personal Information Protection and Electronic Documents Act, S.C. 2000, c.5, Schedule 1, Principle 4.1.3.
29
accountability in the context of synthetic media systems may be diicult and will require evaluation
on a case-by-case basis.
Transparency
Transparency is of particular importance for synthetic media systems. Without this, individuals may
have little understanding of whether and how their personal information is being used. Systems
such as LLMs can produce dierent responses to the same prompt, or responses can change
based on the phrasing of a prompt. How a system generates this content may not be visible or
made public and may constitute a kind of ‘black box’.35
This lack of transparency can make it very diicult to determine whether the system relied on
personal information to inform its response. As well, there may be instances where individuals may
provide their personal information to use a service without knowing that the information could also
be used to train an algorithm or model that may generate content later.
PIPEDA states that organizations must make specific information about their policies and practices
relating to the management of personal information readily available to individuals.36 This means
that information practices must be clear and easy to understand to the average person. For
synthetic media systems, this could mean explaining what personal information is collected, the
purpose for its collection and use (that is, to train a model), how individuals can gain access to their
personal information or complain to the organization, as well as what personal information an
organization discloses to other organizations, and why.
Outputs from a synthetic media system
Issues with the information inputted into a synthetic media system are not the only privacy issues
raised by synthetic media systems. The production of outputs by such a system—for example, an
image, video, text or audio sample—may also engage Canadian privacy laws. The privacy
considerations applicable to such outputs are related to, yet distinct from, those associated with
system inputs.
Definition of personal information
First, organizations must consider whether the generated information is about or related to an
identifiable individual. Under Canadian privacy law, personal information includes any factual or
subjective information, recorded or not, about an identifiable individual.37 For example, if an
organization uses a synthetic media system to generate an image of a specific person, the image
produced would likely be the personal information of the actual individual that the image closely
resembles.
The nature of the technology that creates deepfakes and voice cloning is the creation of something
similar to an original input. If the information used as training data or included in prompts is
35 Snyder, A. (2024). Shedding light on AI’s black box, Axios.
36 Personal Information Protection and Electronic Documents Act, S.C. 2000, c.5, Schedule 1, Principle 4.8.
37 Oice of the Privacy Commissioner of Canada, ‘PIPEDA requirements in brief’, 2024.
30
personal, the generated image, video or audio clip is more likely to also be considered about an
identifiable individual.
The type of synthetic media may also be important to consider. Because the uses of the technology
are broad—it may include anything from text generation to fraud detection, from virtual assistants
to translation services, from summarization to content production and more—the output of LLMs
may vary significantly.
While an LLM may have been trained on personal information, it does not necessarily mean that the
output from the model will constitute personal information itself. There are certainly scenarios
where the content produced by an LLM would not be about an identifiable individual. Thus, whether
the output of an LLM results in personal information, or could lead back to personal information,
needs to be assessed on a case-by-case basis.
Appropriate purposes
Whether the output is considered personal information prompts many other questions. For
example, what are the privacy implications of information that purports to be about an identifiable
individual, but that is ‘fake,’ inaccurate or otherwise unreal?
On the one hand, there may be cases where fake or inaccurate information about an identifiable
individual does not contravene privacy law. For example, synthetic media technologies may be
used to satirize a celebrity or public figure, which under PIPEDA, could potentially be exempt if an
organization’s collection, use or disclosure of the personal information was solely for journalistic,
artistic or literary purposes.
On the other hand, there are real risks that the outputs of a synthetic media system could
themselves contravene privacy law – for example, if the content was purposefully deceptive and
meant to mislead.38 Subsection 5(3) of PIPEDA states that organizations may collect, use and
disclose personal information “only for purposes that a reasonable person would consider
appropriate in the circumstances.” If a deceptive synthetic media product is generated having used
personal information as an input, it’s likely that a reasonable person would consider the use of this
personal information to have been inappropriate, making the use contrary to PIPEDA. This may lead
to several harms, including negatively aecting that individual’s reputation, quality of life or ability
to be in control of their own personal information.
The OPC has highlighted in its guidance certain purposes that will generally be considered
inappropriate. For example, it is inappropriate to collect, use or disclose personal information for
purposes that are known or likely to cause significant harm to an individual.
Significant harm could mean things such as bodily harm, humiliation, damage to reputation or
relationships, and loss of employment, among others.39 In evaluating an organization’s purposes
38 For example, “Nudify Apps” are software applications that use AI to create fake nude images of individuals
without their consent, often generating nude images using non-nude source material. Pendergast, K. (2024).
“Nudify” Apps, the Evolution of Online Exploitation, and What We Can Do About It, Family Online Safety
Institute.
39 Oice of the Privacy Commissioner of Canada, (2018). Guidance on inappropriate data practices:
Interpretation and application of subsection 5(3).
31
under section 5(3), consideration needs to be given to the principles of necessity and
proportionality, including whether the purposes represent a legitimate need and whether the loss of
privacy is proportionate to the benefit.
With this in mind, developers and users of synthetic media systems should be aware of their
purpose for collecting, using and disclosing personal information in having their systems generating
synthetic content, and should assess whether this purpose would be considered, by a reasonable
person, to be appropriate, as required by PIPEDA.
Accuracy and right of access and correction
Principle 4.6 of PIPEDA states that personal information ‘shall be as accurate, complete, and up to
date as is necessary for the purposes for which it is to be used.’40 This raises issues for deepfakes,
where a video or image is communicating false or misleading information about an individual. Such
outputs not only could aect a person’s online reputation but also give rise to other harms related
to misinformation, including around finances, health and public safety,41 and cause individuals to
make decisions based on false or misleading information.
Under PIPEDA, individuals are also entitled to access their personal information, to challenge its
accuracy, and have it amended. Principle 4.9 states that, upon request, an individual shall be
informed of the existence, use and disclosure of his or her personal information and shall be given
access to that information.42
While there are certain exceptions to this access requirement, organizations are expected to be
able to provide people with access to their personal information and to amend it, as required, if an
individual successfully demonstrates that it is incomplete or inaccurate.
For a synthetic media system, this principle may apply to the creator of a synthetic media system,
or to an organization using a particular model. In any event, it may be diicult to retrieve personal
information from these systems, as content may be generated by responses based on machine
learning rather than ‘retrieved from a database.’43 It may require looking at an original data set to see
whether personal information was used, or giving a specific prompt to a model, such as an LLM.44
40 Personal Information Protection and Electronic Documents Act, S.C. 2000, c.5, Schedule 1, Principle 4.6.
41 Digital Regulation Cooperation Forum (2024). The Future of Synthetic Media, p. 21.
42 Personal Information Protection and Electronic Documents Act, S.C. 2000, c.5, Schedule 1, Principle 4.9.
43 Digital Platform Regulators Forum (2024). Examination of technology: Multimodal Foundation Models –
Working Paper.
44 Digital Platform Regulators Forum (2024). Examination of technology: Multimodal Foundation Models –
Working Paper.
32
Final observations
As noted, there are dierent factors to consider when creating, developing, and using a synthetic
media system. While the technology will continue to adapt and change, PIPEDA’s obligations
remain constant and technology neutral. This section has outlined areas where privacy issues and
synthetic media interact, as well as how Canada’s federal private sector privacy law may apply. The
following are some relevant considerations to protecting privacy rights in the context of synthetic
media:
• Synthetic media systems use inputs and produce outputs that may constitute personal
information. In these cases, privacy laws will likely apply.
• Any collection, use or disclosure of personal information related to a synthetic media
system should only be for purposes that a reasonable person would consider appropriate
under the circumstances. For example, generating a video or image which is known or
likely to cause significant harm to an individual would be an inappropriate purpose.
• Organizations must avoid developing or using synthetic media tools that violate ‘no-go
zones’ such as the collection, use, or disclosure of personal information that is otherwise
unlawful. For example, collecting or using the personal information of a person without
their consent to generate nude or sexually intimate images of them would be a ‘no-go
zo ne.’
• Organizations should develop a privacy management program to help correctly identify
privacy-related obligations and risks and appropriately take them into account before they
launch new products or services or implement a synthetic media system.
• Developers and operators should make sure they know the level of sensitivity of the
information being input into a synthetic media system to ensure the appropriate form of
consent and security safeguards used to protect the information. When obtaining consent
from an individual to use their personal information, it should be valid and meaningful.
Security safeguards must be appropriate to the sensitivity of the information.
• The fact that personal information is accessible online does not mean that it can be
collected or used by an organization. Personal information that generally appears to be
publicly available is often still subject to Canadian privacy legislation.
• Organizations should be transparent when developing or using synthetic media -
individuals should know what, how, when, and why their personal information is collected,
used or disclosed throughout any stage of the synthetic media system’s lifecycle. This
should apply throughout a synthetic media system’s development, training and operation.
• Personal information, whether to train synthetic media models, or to be inputted into
them, should be as accurate, complete and up-to-date as necessary for its purpose.
• Generally, individuals have a right to access their personal information as well as the right
to challenge its accuracy and completeness. Organizations should have processes in
place to handle information access requests and ensure the information is accurate.
33
Synthetic Media and Canada’s Anti-Spam Legislation
Canada’s Anti-Spam Legislation’s (CASL) purpose is to build trust in the digital economy by
protecting businesses and consumers from the improper use of technology, including for example,
deceptive marketing practices, spam and other electronic threats.
Together, the CRTC, the Competition Bureau, and the OPC are responsible for promoting
compliance with CASL within a civil regulatory regime.
Commercial electronic messages (Enforced by the CRTC)
CASL prohibits any person—broadly defined to include individuals, legal representatives, and
organizations—from sending commercial electronic messages (CEMs) without consent.
CEMs, including emails, texts, and social media messages, encourage participation in commercial
activities. Such activities include but are not limited to buying, selling, or advertising goods,
services, or investment opportunities. CASL does not apply to political messages, surveys, or CEMs
sent by friends or family.
CEMs must include information about the sender and a way to unsubscribe from receiving further
messages. For more information about when CASL applies, and its requirements visit Frequently
Asked Questions about Canada's Anti-Spam Legislation.
While CEMs are often sent by legitimate companies, they can also come in the form of scams, or
deceptive electronic threats. These can include illegitimate oers, the impersonation of legitimate
companies (for example, banks, government entities), or encouraging the recipient to recover
money or make a payment by clicking a link. When clicked, deceptive links may alter transmission
data or lead to the installation of a computer program without consent, which are also prohibited by
CASL. It is important to highlight that CASL is a civil regulatory regime, and the negative outcome of
these fraudulent activities (for example, stealing personal and financial information) are subject to
criminal laws, not enforced by the CRTC.
In cases of non-compliance, warning letters may be issued, undertakings (an agreement defining
compliance obligations) may be entered into, or notices of violations may be issued, accompanied
by administrative monetary penalties. Individuals may be fined up to $1M and business up to $10M
per violation.
How could commercial electronic messages relate to synthetic
media?
The technologies that produce synthetic media enable persons who send CEMs to do what they
have always done – except it enables them to do it faster, at higher volumes and with less eort.
For example, a common feature of gen AI is its ability to produce text that can then be used for
sending emails, texts or other messages for commercial purposes. Depending on the
circumstances, the CEM must be sent with consent, contain the identity of the sender, and an
34
unsubscribe mechanism. These requirements would have to be met whether the CEM was drafted
by a human or using gen AI.
The Spam Reporting Centre45, where Canadians can report incidents of receiving CEMs without
consent, has received complaints about potentially AI-generated texts and emails; however, it is
not always possible to determine whether a message was in fact drafted using gen AI or was written
by a human. Indicators of an AI-generated email or text may include fewer spelling or grammatical
errors, relatively longer messages and more sophisticated language when compared to past or
other types of messages.46
Legitimate services have policies in place to prohibit misuse of their technologies. However, their
safeguards may be sidestepped by using creative prompts or alternative platforms created for
malicious purposes. Malicious gen AI lack security protections and have been used to
automatically generate convincing emails for phishing purposes.
AI-powered services enable companies to automate email and text generation and send CEMs.
They can also be used to automate marketing chat bots, power semi-autonomous chat bots, as
well as create fake social media accounts that can automatically post content and send direct
CEMs. If the persons behind those electronic accounts are involved in sending CEMs without
consent, they may be in contravention of CASL.
Deceptive marketing practices (Enforced by the Competition
Bureau)
The Competition Act contains civil (section 74.011) and criminal (section 52.01) provisions that
target false or misleading representations and deceptive marketing practices in the electronic
marketplace. These provisions make it illegal to promote a product or business interest, by sending
or causing to be sent false or misleading representations in:
• sender information
• subject matter information and electronic messages
• locator information such as website addresses (URLs)
Additionally, the provisions have the same substantive elements, with the criminal provision having
an added requirement that the representations were made knowingly or recklessly.
How could sections 52.01 and 74.011 apply to synthetic media?
Bad actors may employ deepfakes in electronic messages to promote their schemes. Combining
sophisticated and convincing AI-generated text with the capability to respond to messages
spontaneously due to generative AI technology, email and other electronic message-based scams
can be highly believable and significantly more eective. For example, an employment or
45 To report spam, please see Canada’s Report Spam form.
46 Trustwave (2024). Trustwave SpiderLabs: Artificial Intelligence Playing a Prime Role in BEC and Phishing
Attacks, Trustwave; and Government of Canada (2024). Recognize artificial intelligence (AI): 9 ways to spot AI
content online - Get Cyber Safe.
35
investment opportunity that is promoted via electronic messages, such as emails and direct
messages, may contain any of the following:
• Unauthorized deepfake logo of a well-known corporation
• Unauthorized deepfake video of well-known public figures falsely endorsing the opportunity
• Convincing text messages that appear legitimate
In return, however, consumers are asked to pay a fee or are required to provide personal
information to access the opportunity.
How does unauthorized collection and use of personal information
relate to synthetic media? (Enforced by the OPC)
Through CASL amendments to PIPEDA, the OPC focuses on enforcing and investigating the
unauthorized collection and use of personal information obtained:
• Through the harvesting of electronic addresses (for example, email accounts, text accounts
and social media instant messaging accounts), where lists of such addresses are compiled
through mechanisms that include the use of computer programs that automatically mine
the Internet, or
• Through the gaining of unlawful access to individuals’ computer systems, primarily through
the use of malware/spyware.
Address harvesting refers to various techniques used to automatically compile or generate lists of
electronic addresses for bulk electronic mailings. This can be done by spammers themselves, or
other entities (such as electronic address harvesters) who then sell the address lists.
The collection of personal information through illicit access to other people’s computer systems
includes computer programs such as malware or spyware that collect personal information and are
downloaded and remotely installed on a computer without the user’s knowledge.
As noted above, the installation of computer programs without consent is prohibited by CASL. Such
programs can be used, among other things, to covertly gather information about an individual’s
web-browsing, computer activities, user credentials or financial information via “keylogging”, or
scraping electronic addresses from address books or instant messaging contacts. For more
information, see the OPC's webpage on responsibilities under CASL.
Gen AI using synthetic data can potentially be used by spammers, both commercial or malicious,
to harvest electronic addresses in a more eicient and cost-eective manner. This could make it
easier for new spammers to start harvesting, and for existing spammers to expand their activities.
The use of synthetic media to create and deliver more persuasive commercial or malicious
electronic communications will inevitably lead to spam campaigns gaining a higher participation
rate by targeted individuals. This may be done by:
• increased click rates on links to malicious apps;
• spoofed websites that have a more professional appearance; or
36
• interactions with fake chatbots that appear to be other humans.
This could result in greater success for those seeking to collect and use personal information for
commercial or malicious purposes, than was potentially achievable without the use of AI.
In 2023-2024, after publishing a Joint Statement on Data Scraping and the Protection of Privacy, the
OPC and eleven other international data protection authorities engaged with several global social
media companies. The OPC and its counterparts reminded companies of their obligations to
protect users’ personal information from unauthorized access and data scraping.
The companies informed the OPC that scrapers were using AI to scrape data more eectively (for
example, via “intelligent” bots that can simulate real user activity). At the same time and on a
positive note, companies explained that they too were employing AI to better detect and protect
their platforms and users against unauthorized scraping, highlighting that innovative AI tools can
also be part of the solution. See the concluding Joint Statement on data scraping and the protection
of privacy.
Conclusion and key takeaways
AI-related emerging technology, such as gen AI, is an area of interest to many—consumers,
businesses, and government authorities. While the growth in popularity and advancement in gen AI
technology brings a spectrum of benefits, they also come with potential risks. This paper explored
some of the potential implications of synthetic media.
Synthetic media may revolutionize the Canadian economy by enabling innovative content creation,
enhancing customers’ experiences, and improving operational eiciencies. Many sectors of the
Canadian economy can benefit from the use of gen AI which enables synthetic media to be easily
and quickly produced at a lower cost. For example, the creative industries may be able to reduce
production costs and save time by using gen AI to assist with tasks such as scriptwriting, dubbing,
and special eects. Synthetic media can also enhance the accessibility of media content and
services in a variety of ways.
However, synthetic content may pose threats to privacy, security and trust in public institutions and
the news. Synthetic media may be misused to facilitate deceptive marketing practices. Other
potential issues include the collection, and use of personal information to train algorithms or
generate content without consent, as well as risks to people’s reputation, and the spread of
misinformation and disinformation. The legal uncertainties surrounding copyright and authorship,
as well as the economic consequences of job displacement, are also some potential risks.
Key Takeaways
Throughout this report, the Members highlighted benefits and risks that may be associated with
recent developments in the world of gen AI and synthetic media, summarizing the potential
implications. The Members have highlighted three consistent themes:
37
• Truthfulness and transparency: Companies developing or using synthetic media should
be truthful and transparent with their users and consumers, as Canadians might want to
know when gen AI was used in the creation, oering and advertising of a product or service.
Transparency can also help determine whether an organization is using copyrighted
material.
• Staying informed: Continued Member collaboration will assist in keeping pace with the
evolution of synthetic media and its impact on various sectors of the Canadian economy.
Engaging with experts, industry stakeholders, consumers, and other governmental agencies
can aid in building domestic capacity while addressing challenges.
• International collaboration and coordination: Given the global nature of these emerging
technologies, international collaboration is crucial to ensuring that policymakers are kept
abreast of the most recent developments and can coordinate their eorts. In this sense, the
CDRF is well positioned to leverage its membership in the International Network for Digital
Regulation Cooperation to closely collaborate with its counterparts.
While synthetic media may benefit Canadians, it is also likely to require adequate monitoring to
further its potential benefits and mitigate its potential risks. By embracing transparency and
fostering both domestic and international cooperation, the Members can collectively navigate the
evolving digital landscape to serve the interests of Canadians.
