In November 2024, [several Canadian media companies sued OpenAI](https://litigate.com/assets/uploads/Canadian-News-Media-Companies-v-OpenAI.pdf), alleging it circumvented technological protections to scrape their sites and copied their works, infringing copyright. This lawsuit and [other ongoing legal battles]( https://nytco-assets.nytimes.com/2023/12/NYT_Complaint_Dec2023.pdf) underscore the growing tension between AI technology and existing copyright law, as well as the evidentiary challenges inherent in large-scale data-scraping claims. 

Despite alleging massive infringement, the suit offers little concrete evidence, raising the question of whether broad accusations can compel an AI company to open its “black box” without a firmer factual basis. While these cases highlight legitimate concerns about using copyrighted material in AI training, they rest largely on speculation and circumstantial inferences. Indeed, the Canadian media companies are effectively asking OpenAI to prove a negative: to show it neither scraped their sites nor relied on their data for training. 

The claims may be plausible, but outside of the confidential walls of OpenAI, they appear to lack evidentiary causal links to damages: details of circumvented protections and clear proof of copying. As filed, their chance of success appears uncertain, and the independent evidence suggested by the pleadings alone likely falls short of what is needed to prevail at trial. 

### The Problem of Circumstantial Evidence

Many of the claims against OpenAI (and other AI developers) rely heavily on inference, assumptions, and circumstantial evidence. The Canadian media companies’ arguments hinge on a few key points: 

**The "Likely Included" Argument:** The plaintiffs state that because their works were accessible on the internet, and AI models are trained using vast amounts of online data, their works were likely included in training datasets. However, the statement of claim also acknowledges that this is assumed and not known: “The full particulars of when, from where, and exactly how, the Works were accessed, scraped, and/or copied is within the knowledge of OpenAI and not the News Media Companies.” 

 The pleadings refer broadly to OpenAI “scraping” millions of articles from the media companies’ sites, or from authorized third parties. But the plaintiffs have not identified specific articles or excerpts, nor provided examples of ChatGPT outputs that replicate or closely paraphrase those articles. Copyright infringement requires proof of copying a work (or a substantial part of it). Alleging that “you copied an enormous number of articles” typically requires at least some reference to a sample set of works or examples of outputs that reproduce distinctive content. This is notably absent from the Canadian media companies’ statement of claim. 
 
**Circumvention and Scraping Allegations:** The plaintiffs also allege that OpenAI “scraped” their websites, and that since “paywalls” or “robots.txt” files were in place to prevent scraping, OpenAI must have bypassed those controls.  

Yet there are no concrete facts explaining how they know OpenAI circumvented these restrictions, nor any technical details of detected intrusions. Stating the conclusion that “OpenAI circumvented our technological measures” is different from pleading the material facts establishing circumvention. 

**Model Outputs as "Proof":** Although not specifically pled by the Canadian media companies, other plaintiffs point to the ability of AI models to reproduce outputs that are similar to their works as evidence of "memorization" and, therefore, of inclusion in the training data. For instance, the New York Times claims that GPT-4 can produce near-verbatim excerpts of some of their articles, although this alone does not prove direct, unauthorized use of their content.  

The Canadian media companies make a similar claim alleging OpenAI broadcast, displayed, transmitted, and published their works. But unlike the NY Times, the Canadian media companies do not include any examples of this occurring, and their statement of claim is vague as to whether they have any evidence in hand. 

### The Lack of Specifics: A Trial of Vague Claims

These arguments share a common thread: they demonstrate that the media companies do not possess evidence that specific content was used, where it was obtained, or how it was used within the training process. Instead, they suggest it is likely that this was the case. They rely on a combination of availability (content was available on the internet), common practices (AI models are trained on vast amounts of data including scraped content), and output similarity (outputs are sometimes similar to plaintiffs’ works). These points, while suggestive, fall short of the burden of proof typically required in copyright infringement cases unless they can be substantiated by direct evidence. 

Beyond the circumstantial nature of the evidence, the claims often suffer from a distinct lack of specificity. Consider these issues: 

**No Specific Datasets:** Plaintiffs rarely identify specific datasets that contain their works. They often refer to large, publicly available datasets without naming the precise datasets or their composition. In the New York Times and Canadian media companies’ lawsuits, they mention datasets like Common Crawl (an open dataset of web data), but the lawsuit does not specify which content came from which sources. Notably, there is no suggestion that the Common Crawl dataset is infringing; only its use by OpenAI is alleged as unauthorized copying. 

**Vague Allegations of Copying:** The Canadian media companies refer to their millions of works and infer that OpenAI copied and ingested them without concrete details about how, when, or from where this occurred. The claim states that OpenAI scraped their content directly from their websites, or from third-party datasets, but does not clarify which content was taken from each source. 

**Unclear or No Real-World User Infringement:** Many claims fail to clearly explain how the use of the content constitutes copyright infringement. They claim the models were trained using their content and/or can reproduce their content, but do not show how this breaks any copyright law or results in harm. 

In the New York Times lawsuit, the claims cite examples where the AI model was prompted with specific article excerpts and output similar text. However, these prompts were created by the plaintiffs, not typical users. There are no allegations that real-world users are doing this, or why such behaviour would cause harm. 

The Canadian media companies allege that by making its models available, OpenAI effectively authorized copyright infringement. However, it remains unclear whether they contend that OpenAI’s conduct meets the Canadian legal standard of authorization — that it expressly or implicitly sanctioned, approved, or countenanced the infringing activity — or whether they are arguing that merely providing the means to infringe should be sufficient to establish authorization. The burden of proof is on the plaintiff to show that the defendant intended to authorize the infringing activity, and [this cannot be assumed just because the defendant has the ability to prevent it](https://www.canlii.org/en/ca/scc/doc/2004/2004scc13/2004scc13.html#par42). This will be difficult to prove considering the efforts undertaken by [OpenAI to specifically prevent verbatim outputs and memorization](http://arxiv.org/abs/2412.06370). 

### The Problem of Proving a Negative

At the pleading stage, plaintiffs are not required to give every factual detail necessary to substantiate a legal claim, especially when the details are in the defendant’s control. The Statement of Claim appears to plead the material facts necessary to support its causes of action, but the pleadings at times mix factual allegations with inferences and conclusions. Whether these allegations will hold up if challenged at trial is a separate question that depends on whether the plaintiffs can produce actual evidence in discovery to prove those facts: a few ‘smoking-gun’ examples, or server logs, or AI outputs that strongly indicate copying.  

This lack of specifics makes it difficult to establish a clear case of copyright infringement. It also creates a situation where the plaintiffs are, in effect, asking AI companies to prove a negative: to prove that their data was not used. This approach challenges the established legal principle that the burden of proof rests with the plaintiff. 

Demonstrating that specific data was not used is a difficult, if not impossible, task. It is much easier to demonstrate that an action did occur than to prove that an action did not occur. Because training datasets are complex and often opaque, the plaintiffs risk shifting the burden of proof, demanding that OpenAI demonstrate that the plaintiff's content was not present, or was not used in an infringing way. 

This situation creates an unfair imbalance. The plaintiffs, by relying on the claim of "likely" inclusion and circumstantial evidence, are essentially asking the court to compel OpenAI to disclose their training data records, but without first providing specific evidence of copyright infringement. 

### Implications and Questions

This situation raises critical questions about how AI training applies to copyright law. The lawsuits against OpenAI highlight the challenges of litigating copyright issues related to AI, particularly due to the opaqueness of training datasets and the difficulty for copyright holders to obtain direct evidence of infringement. 

The Canadian media companies must make a more compelling case based on demonstrable facts rather than broad assertions and circumstantial arguments. The statement of claim repeatedly states that the particulars are “… solely within the knowledge and control of the OpenAI Entities and not the News Media Companies,” implying the plaintiffs lack the evidence to support their claims. Unlike the New York Times claim that includes examples of specific infringement, the Canadian media companies rely on inferences, assumptions, and circumstantial evidence to support their vague claims.  

To be successful, the Canadian media companies will have to adduce evidence to support their claims: log or forensic evidence of scraping and clear proof of circumventing technological protection measures (TPM); specific examples of copied files (relying on outputs); and evidence verifying that training sets included plaintiff’s articles.

While the media companies’ statement of claim likely pleads the material facts necessary to support the cause of action, there appears to be evidentiary gaps supporting these facts. Relying on OpenAI to fill these gaps in discovery is risky. 

As litigation unfolds, courts may articulate new legal tests or processes for AI-based copyright claims that could guide both future plaintiffs and AI developers on how to plead, defend, or avoid allegations of large-scale data scraping. Legislatures worldwide may also intervene as [several countries consider allowing text and data mining for training AI]( http://data.europa.eu/eli/reg/2024/1689/oj/eng).  Ultimately, these lawsuits will test whether broad, circumstantial allegations can compel AI developers to disclose their entire data pipeline, or whether courts will insist on stronger initial proof. Whichever way it goes, the result could shape AI copyright jurisprudence for years to come. 

_The opinion is the author's, and does not necessarily reflect CIPPIC's policy position._

Commentary

medium_Untitled design (1).png

small_Untitled design (1).png

thumbnail_Untitled design (1).png

Untitled design (1).png

The AI Copyright Conundrum: Are Claims Against OpenAI Built on Sand? 

Victims of non-consensual distribution of intimate images (NCDII) deserve quick takedown methods to stop the spread of the images while they decide whether to mount further legal action. Yet, due to online services’ self-regulation of NCDII, takedown responsiveness and routes vary from one online service to the next.  

Two main takedown routes address NCDII, both grounded in consent and a harms-based framing: copyright and self-regulated NCDII takedown processes. Copyright recognizes harm in sharing someone’s intellectual property (the image itself, the photo or video) without consent. Copyright prioritizes authors’ exclusive rights and the profit-potential of their work. NCDII processes, in contrast, recognize harm in the denial of privacy and bodily autonomy. The focus shifts from profit to consent for publishing a person’s image – a priority not well-supported by current Canadian law. 

Some online services have voluntarily implemented NCDII-specific takedown tools in the absence of government regulation. But, in the case of X.com, those self-regulated takedown tools may be just for show. 

Researchers from the University of Michigan and Florida International University posted [50 AI-generated nudes to X.com](https://arxiv.org/pdf/2409.12138). 25 were reported to X by the researchers as copyright infringement, and 25 as non-consensual nudity under the site’s own [non-consensual nudity policy](https://help.x.com/en/rules-and-policies/intimate-media). After 25 hours, X removed all of the copyright-reported intimate images. However, the platform removed none of the images reported as containing non-consensual nudity by the end of the study’s 3-week observation window.  

These results suggest that copyright is a quick and easy solution to NCDII. However, copyright law’s ability to address harm has clear limits. 

The imbalance instead suggests there is space for regulatory intervention to encourage quick action on NCDII reports. Specifically, through some obligation for online services to act quickly on reports of NCDII on their platforms. 

#### Copyright has its limits 

Copyright is a legal right that arises as soon as someone creates a literary, dramatic, musical or artistic work. It focuses on harms to the work itself. When I take a selfie, I automatically hold copyright in that image. If someone else gets their hands on that selfie and posts it publicly without my consent (authorization), they are likely infringing the copyright I own in that image. At that point, I can file a copyright takedown notice through the host service (Facebook, Instagram, X, etc.) to have it removed.  

But copyright takedown requests go exactly as far as that. The material is taken down and repercussions generally end there, with no acknowledgement of the [criminal nature of NCDII](https://laws-lois.justice.gc.ca/eng/acts/C-46/section-162.1.html), no follow-up with the perpetrator, no acknowledgement of harm to the individual, and no guarantee the material won’t be reshared elsewhere. and no guarantee the material won’t be reshared elsewhere.   

What’s more, a victim of NCDII may not be the copyright holder. If a partner, friend, or even a stranger snapped the photo or recorded the video, that person almost inevitably holds the copyright. This means that a platform would have no basis in copyright law to take down any image of a person taken by a third party, including what are referred to as “creepshots”, “upskirts.” Nor would a platform need to act on images taken by a victim’s partner or friend – consensually or not.  

Copyright’s shortcomings point to a need for takedown routes for victims that do not hold copyright, and that acknowledge the harms of NCDII. NCDII-specific reporting tools may be able to fill this gap, but need to be properly resourced to work. 

#### NCDII reporting tools rely on online services’ pinky swear to self-regulate

Reporting tools have their limits, too. Governments have historically pressured online services to self-regulate harmful content. The concept is that trust and safety are an online service’s purview and users will reward good self-regulation by flocking to more reputable services. As such, large online services that allow user uploads generally have an NCDII policy as part of their trust and safety programs. But the takedown systems themselves are not a silver bullet and rely on online services to actually self-regulate. 

X.com’s self-regulation-based NCDII system saw [no attention at all](https://arxiv.org/pdf/2409.12138) given to reported intimate images after 3 weeks. While trust and safety were thought to be central to an online service’s success, X seems to have thrown trust out the window. The company even renamed its Trust and Safety department to simply “Safety” [in the spring of 2024](https://www.cnn.com/2024/04/02/tech/elon-musk-x-new-head-of-safety) after Musk’s 2022 firing of [80% of the department's staff](https://techcrunch.com/2024/09/04/x-is-hiring-staff-for-security-and-safety-after-two-years-of-layoffs/). Clearly, at least some online services are de-emphasizing self-regulation. So can we make online services actually self-regulate? 

A prime example of consequences for a lack of self-regulation comes from Pornhub, which hosted vast swathes of user-uploaded NCDII and child sexual abuse material (CSAM). Its lack of adequate content moderation allowed harmful content to flourish. After [public outcry](https://nationalpost.com/news/canada/remove-non-consensual-content-enforce-existing-laws-ethics-committee-on-pornhub), [costly lawsuits](https://globalnews.ca/news/8286778/pornhub-lawsuit-settlement/), and Visa and Mastercard’s decision to [withdraw payment processing services](https://www.reuters.com/business/finance/mastercard-visa-suspend-ties-with-ad-arm-pornhub-owner-mindgeek-2022-08-04/), the company implemented a comprehensive trust and safety program. This resulted in total CSAM reporting on the site from over 13,000 in [2020](https://www.missingkids.org/content/dam/missingkids/pdfs/2020-reports-by-esp.pdf), down to three in [2021](https://www.missingkids.org/content/dam/missingkids/pdfs/2021-notifications-by-ncmec-per-esp.pdf), two in [2022](https://www.missingkids.org/content/dam/missingkids/pdfs/2022-notifications-by-ncmec-per-esp.pdf), and zero in [2023](https://www.missingkids.org/content/dam/missingkids/pdfs/2023-notifications-by-ncmec-to-responsive-esp.pdf) with a takedown timeframe of .2 days. 

If self-regulation is actually going to work, imposing consequences for not doing it may be useful. Preferably consequences that don’t require years of legal action by victims to realize. 

#### How can we make this experience better for victims, and even prevent NCDII?

Important lessons abound in the legislative and policy frameworks that have prioritized consent to use of intellectual property over personal privacy and bodily autonomy. Publishing intimate images without consent of the person in the images is even illegal under the [Criminal Code of Canada](https://laws-lois.justice.gc.ca/eng/acts/C-46/section-162.1.html). But the process of criminal prosecution of perpetrators can be traumatic and lengthy for victims. With limited liability or other legal repercussions for what users post, services like X have little incentive to move quickly on complaints of non-consensual intimate image sharing.  

#### So, what can we do?

One possible solution comes in Canada’s Online Harms Act, though the Act is unlikely to pass in the vestiges of the 44th Parliament. The Act would introduce [a duty of care](https://www.osler.com/en/insights/updates/canada-s-new-online-harms-act-c-63-what-you-need-to-know/) for online services to take down material reported as NCDII. Any flag of NCDII that is not “trivial, frivolous, vexatious or made in bad faith” would trigger a duty for the online service to make the content unavailable to the public. Importantly, the Act requires online services to retain the material on record - making space for criminal investigations or dispute processes. This framework maintains the service’s ability to host user-uploaded content, only triggering liability if they fail to act within 24 hours. 

Quebec’s [Bill 73](https://assnat.qc.ca/en/travaux-parlementaires/projets-loi/projet-loi-73-43-1.html), An Act to counter non-consensual sharing of intimate images, takes a similar route. The Act imposes fines or imprisonment on any person that does not act on a court order to take down NCDII.  

Europe has implemented the General Data Protection Regulations (GDPR), a privacy-centered directive that considers images of individuals as their own personal data, no matter who created the image. This creates a sort of [right to one’s image](https://www.lexology.com/library/detail.aspx?g=f9882cc1-c24c-4313-b986-655610b03d6d) that roughly mirrors the right to a work under copyright. That said, [NCDII has not evaporated in Europe](https://www.politico.eu/article/how-europe-privacy-laws-are-failing-victims-of-sexual-abuse/). It remains difficult to enforce privacy provisions because of a lack of supporting takedown timelines or requirements like those provided for under the Digital Millenium Copyright Act (DMCA).  

The DMCA is American legislation that shields online services from copyright liability if they act promptly to DMCA takedown requests. If a copyright owner puts in a request for an online service to take down copyrighted material that appears on the service without the owner’s consent, the DMCA requires the online service remove the content. This process is called notice and takedown. Applying a notice and takedown approach to a GDPR-like protection of personal data in images, could create the quick-response model that NCDII requires. 

But liability-based systems aren’t perfect. There are risks of misuse of mandatory-action takedown systems to ban adult content online - even adult content produced with full consent. There’s also a chance that platforms will simply amplify content filtering to filter possible NCDII before it’s posted, disproportionately affecting women’s expression as their content is often flagged as sexual [even when it is not](https://www.theguardian.com/technology/2023/feb/08/biased-ai-algorithms-racy-women-bodies). Further, introducing liability cuts into online services’ bottom line, requiring more content moderation staff, who often work in [conditions likened to torture](https://apnews.com/article/kenya-facebook-content-moderation-lawsuit-8215445b191fce9df4ebe35183d8b322) abroad. 

NCDII is tech-facilitated sexual abuse that can proliferate at alarming rates as users share, download, and re-share content. It requires action. But that action must be thoughtful, targeted and informed by the social realities and conceptions of consent and bodily autonomy that underpin NCDII.  

Themes of power and control emerge in NCDII, centered in the reality that society demonizes sexual women and 2SLGBTQ+ folks. The notion that the victim may have [“asked for it”](https://inside.southernct.edu/sexual-misconduct/facts) by even allowing the image to be taken in the first place creates space for online services not to act, redirecting blame to the victim and protecting the perpetrator. Continued shunning of victims again punishes NCDII victims themselves for their supposed “bad behaviour” while rewarding the perpetrator with the result they desired – harming their victim. We, collectively, must let go of these myths to undercut the power that motivates NCDII before it even happens.   

Indeed, we need many tools in the toolbox to fight against NCDII. And it starts with prioritizing bodily autonomy and privacy, believing victims, and collectively saying that enough is enough. 

_The opinion is the author's, and does not necessarily reflect CIPPIC's policy position._

 
**If you have been affected by non-consensual intimate image sharing, Tech Safety Canada operates [a useful help document](https://techsafety.ca/resources/toolkits/what-to-do-if-someone-has-shared-or-threatened-to-share-your-intimate-image). It includes instructions on navigating popular platforms’ reporting tools, emotional supports, and legal avenues.**


medium_X Marks the Spot blog image.png

small_X Marks the Spot blog image.png

thumbnail_X Marks the Spot blog image.png

X Marks the Spot blog image.png

X Marks the Spot: Where Copyright Wins, Consent Loses, and Victims Get Left Behind 

A few years ago, open-source artificial intelligence (AI) models [dominated](https://github.com/readme/guides/open-source-machine-learning) [conversations](https://www.brookings.edu/articles/five-ways-that-open-source-software-shapes-ai-policy/) about AI. This vision for a democratized and collaborative AI ecosystem seemed bright. Fast forward to today and the contributions of open-source models are overshadowed and undervalued. Big tech players are monopolizing the AI industry, replaying a story the technology industry knows all too well. 

Big tech controls AI development by leveraging existing advantages surrounding [access to data, compute power, and the talent](https://competition-policy.ec.europa.eu/document/download/79948846-4605-4c3a-94a6-044e344acc33_en?filename=20240723_competition_in_generative_AI_joint_statement_COMP-CMA-DOJ-FTC.pdf) needed to run large-scale AI models. While open-source AI models make important contributions, the changing regulatory and industry environment threatens and undervalues the critical role open-source models play in the industry. Conversations on AI have shifted from fostering innovation to addressing safety, controlling risk, and regulating AI. As governments around the world grapple with how to regulate AI, open-source models must be re-centred in this discussion to grow a competitive, transparent, and innovative AI industry. 

#### Why open-source AI is critical
Supporting open-source AI is key to creating a thriving AI ecosystem. Open-source models ensure [more people have the opportunity to use AI and get involved with its development](https://cset.georgetown.edu/article/open-foundation-models-implications-of-contemporary-artificial-intelligence/), without relying on a few big tech companies. This creates a more dynamic and competitive market, with several overall benefits. 

**Transparency and Accountability:** Open code allows others to inspect and audit AI models for biases and errors. For example, [BLOOM is a multilingual model collaboratively developed by over 1,000 researchers worldwide](https://www.nature.com/articles/d41586-022-01705-z).This open approach has allowed several researchers to improve BLOOM over time. 

**Cost and Flexibility:** Open-source models [cut costs for organizations](https://datasciencedojo.com/blog/open-source-llms-for-enterprises-benefits/) looking to apply AI to their business. Businesses can also customize these models to meet specific needs, which is especially beneficial for small and medium-sized enterprises looking for affordable AI integration.

**Safety, Security and Privacy:** Open-source AI promotes safety and security because more people are [reviewing the technology to fix biases, vulnerabilities, and other flaws](https://cset.georgetown.edu/article/open-foundation-models-implications-of-contemporary-artificial-intelligence/). Organizations can also [maintain control over data by hosting their own version](https://datasciencedojo.com/blog/open-source-llms-for-enterprises-benefits/) of an open model. Using closed models often requires companies to send sensitive data to third-party servers, which increases the risk of data breaches. 

#### Subtle shifts from open to closed AI
Conversations about AI often interchangeably use terms like open-source models, open models, and open weights, without definition. Open-source AI refers to models which [make the model’s code and training data openly available](https://cset.georgetown.edu/article/open-foundation-models-implications-of-contemporary-artificial-intelligence/). AI model code shapes the entire model by processing data, defining the model’s architecture, training and evaluating the model, and facilitating deployment for use. Others can use, model, and improve the AI model when the code is available. 

Open models can [vary in the degree of openness](https://cset.georgetown.edu/article/open-foundation-models-implications-of-contemporary-artificial-intelligence/). A model’s openness is influenced by factors like [access to training data, availability of documentation on data collection and processing, and transparency about model design and testing](https://cset.georgetown.edu/article/open-foundation-models-implications-of-contemporary-artificial-intelligence/). The [Open Source Initiative](https://opensource.org/definition-annotated) recently released a definition of open-source AI. Under this [new definition](https://opensource.org/ai/open-source-ai-definition), an open-source AI is a system that allows users to freely use, study, modify, and share the system without permission. An AI system [qualifies as “open”](https://opensource.org/definition-annotated) if the system includes the preferred form for modifying training data, source code, and model parameters. While the definition of open-source AI is [still debated](https://techcrunch.com/2024/10/28/we-finally-have-an-official-definition-for-open-source-ai/), this definition provides a solid basis for evaluating a model’s openness. 

Meta’s LLaMA and BLOOM show varying degrees of openness. Meta discloses its LLaMA model weights, but it does not qualify as a truly open-source model under the [Open Source Initiative’s definition](https://opensource.org/ai/open-source-ai-definition) because it imposes restrictions on commercial use. BLOOM is lauded as an example of true open-source AI because [every element of the model is freely accessible for inspection and further improvement](https://dailyai.com/2024/06/the-rise-of-open-source-ai-models-transparency-and-accountability-in-question/).

Closed-source models, like OpenAI's ChatGPT or Google's PaLM, are  proprietary and controlled by developers. The [model’s source code is not made available for others to use or audit](https://www.weforum.org/agenda/2023/12/ai-regulation-open-source/).

The shift from open to closed AI models is already underway. For example, OpenAI previously published information on its earlier ChatGPT models. However, when launching GPT-4, Open AI announced it would not publish any details about the [“architecture, hardware, training compute, dataset construction, training method, or similar,” citing the competitive AI landscape and safety implications](https://ainowinstitute.org/wp-content/uploads/2023/04/AI-Now-2023-Landscape-Report-FINAL.pdf).

#### The future for open-source AI
The rising dominance of large, closed models has coincided with AI regulation, which threatens to curtail the vital contributions open-source models are making to the ecosystem. Governments must balance the unique role of open-source models in driving innovation with regulatory compliance burdens. If compliance becomes too burdensome, especially for small developers, it could stifle AI innovation.

New AI regulations are already struggling to strike this balance. The European Union’s AI Act provides limited exceptions for open-source models, which will make it [difficult for individual or small developers to meet the Act’s compliance burden and bring their models to the EU](https://datainnovation.org/2024/03/the-eus-ai-act-creates-regulatory-complexity-for-open-source-ai/). This time last year, [President Biden’s executive order on AI regulation](https://www.whitehouse.gov/briefing-room/presidential-actions/2023/10/30/executive-order-on-the-safe-secure-and-trustworthy-development-and-use-of-artificial-intelligence/) sparked debate among open-source advocates. The order defines “dual-use models” as AI models trained on broad data which “generally use self-supervision.” Organizations like [Y Combinator](https://x.com/martin_casado/status/1720517026538778657) argued these overly broad definitions will capture most of the AI industry and place undue restrictions on small businesses and open-source models. As governments worldwide contemplate AI regulation, finding the right balance will be crucial to preserving an innovative and competitive AI industry.

Open-source models already have and continue to make powerful contributions to the AI ecosystem. In this critical moment for AI development, policymakers must recognize and protect the contributions open-source models are making to this ecosystem. The open-source movement is about democratizing the future of technology. We need to keep this vision for open-source AI alive and prevent it from fading in a closed, monopolized AI market.

_The opinion is the author's, and does not necessarily reflect CIPPIC's policy position._


small_Picture1.png

thumbnail_Picture1.png

Picture1.png

Bolstering Open-Source AI in the Era of Closed, Big Tech Models  

Commentary / Blogs

Perspectives

Les programmes de stage de CIPPIC offrent aux étudiants la possibilité de travailler avec des avocats expérimentés sur des problèmes juridiques réels et de pointe résultant de l'utilisation de nouvelles technologies

Travail étudiant

Thèmes

CIPPIC has filed its [Memorandum of Fact and Law](https://cms.cippic.ca/uploads/A_267_24_1395804_ONTARIO_LTD_v_AG_Canada_Intervener_Memorandum_of_Fact_and_Law_3e8eead498.pdf) in the appeal of [_1395804 Ontario Ltd. v. Canada (Attorney General_), 2024 FC 829]( https://canlii.ca/t/k4zfr), arguing that the _Copyright Act_’s anti-circumvention provisions do not override fair dealing. CIPPIC’s argument focuses on reconciling user rights with Parliament’s enactment of remedies for circumventing digital locks. This intervention follows CIPPIC’s successful participation in the hearing in the Court below, where CIPPIC’s interpretation of the digital lock provisions prevailed. 


**The Decision: Fair Dealing and Anti-Circumvention Provisions**

In the decision under appeal, [_1395804 Ontario Ltd. v. Canada (Attorney General)_, 2024 FC 829]( https://canlii.ca/t/k4zfr), the Federal Court of Canada ruled that Canada’s anti-circumvention provisions do not override fair dealing. The plaintiff, a paywalled news service, sued Parks Canada, arguing that employees had improperly accessed and shared its articles, violating copyright and technological protection measures. The court rejected this claim, holding that using a validly obtained password is not circumvention and, regardless, copyright owners cannot use digital locks to negate fair dealing rights.
The unsuccessful plaintiff appealed.

**CIPPIC’s Argument on Appeal**

CIPPIC has been granted leave to intervene in the appeal. It has intervened to defend the balanced interpretation of the relationship between user and digital locks in copyright law. We argued that:
- Anti-circumvention provisions must align with fair dealing. Copyright owners cannot use technological protection measures to eliminate legally protected user rights.
- A valid password is not circumvention. The Copyright Act targets hacking, not legitimate access.
- The Act must remain technologically neutral. A user’s fair dealing rights do not disappear when content is digital rather than physical.

CIPPIC thanks [James Plotkin](https://gowlingwlg.com/en-ca/people/james-plotkin) and his team at Gowling WLG for their outstanding work on our intervention.

**Next Steps**

The Federal Court of Appeal will schedule a hearing. CIPPIC will continue defending fair dealing and ensuring technological protection measures do not become tools for erasing user rights.
- CIPPIC’s Memorandum of Fact and Law](https://cms.cippic.ca/uploads/A_267_24_1395804_ONTARIO_LTD_v_AG_Canada_Intervener_Memorandum_of_Fact_and_Law_3e8eead498.pdf)
- [Intervention Order](https://cms.cippic.ca/uploads/A_267_24_20241220_O_E_O_OTT_20241220091845_LB_2_c79d6cd3bf.pdf)
- Decision under review [_1395804 Ontario Ltd. v. Canada (Attorney General)_, 2024 FC 829](https://canlii.ca/t/k4zfr)	

Fair Dealing and Digital Locks: CIPPIC’s Intervention at the FCA

On February 14, 2024, CIPPIC submitted an intervention on behalf of [OpenMedia](https://openmedia.org/) in [Telecom Notice of Consultation CRTC 2024-293](https://crtc.gc.ca/eng/archive/2024/2024-293.htm) to advocate for enhanced transparency and fairness in Canada’s telecom market.

The submission was inspired by Ontario Commissioner Bram Abramson’s dissenting opinion, which argued that the current consultation misses key opportunities to enhance consumer information. Commissioner Abramson advocated for a more comprehensive, data-driven approach to consumer protection. CIPPIC was glad to support OpenMedia’s position that the CRTC’s current approach does not go far enough in addressing the transparency issues that prevent Canadians from making informed decisions about their mobile wireless services.

In a [parallel proceeding](https://crtc.gc.ca/eng/archive/2024/2024-318.htm), the CRTC is considering “nutrition labels” to help Canadians better understand and compare Internet service plans. However, the proposed labels do not include mobile wireless services—a significant oversight given that mobile consumers face the same challenges as Internet users. Confusing and inconsistent plan information leaves many Canadians struggling to make informed choices. Worse, mobile wireless complaints accounted for over half of all telecom complaints last year, highlighting the urgent need for better transparency.

CIPPIC’s submission on behalf of OpenMedia calls on the CRTC to expand standardized “nutrition labels” to include mobile wireless services. By adopting clear, machine-readable labels, the Commission can empower consumers, reduce confusion, and foster a more competitive telecom marketplace. Labelling should not stop at service plans. Consumers deserve transparency about the environmental impact and durability of the mobile devices they purchase.


Read the full submission [here](https://cms.cippic.ca/uploads/Open_Media_Intervention_CRTC_2024_293_3ed7018aeb.pdf).

Championing Consumer Rights: OpenMedia’s Submission on Telecom Transparency

CIPPIC is excited to announce a renewed focus on its open government work through its new Open Information Initiative. With the next mandated review of Canada's _Access to Information Act_ commencing by June 2025, CIPPIC hopes to draw attention to the worrying state of access to information in Canada.

During this time, CIPPIC will be sharing the results of its access to information requests with journalists working in open-access publications to emphasize the importance of access to information-driven journalism. CIPPIC will also publish its own short articles, draft policy briefs, partner with stakeholders, comment on transparency issues, and publish a comprehensive report on the current system.

Contact: admin@cippic.ca.

***

Some of this work: 

- Elizabeth Thompson, "[Violent extremists are using antisemitism to recruit in Canada: CSIS report](https://www.cbc.ca/news/politics/csis-antisemitism-violent-extremists-1.7463398)," CBC, Feb. 20, 2025. 
- Bill Curry and Mahima Singh, "[Ottawa spent record amount on outsourcing despite vow to rein in practice](https://www.theglobeandmail.com/politics/article-ottawa-spent-record-amount-on-outsourcing-despite-vow-to-rein-in/)," Globe and Mail, Feb. 18, 2025. 

The Launch of CIPPIC's Open Information Initiative

#### Prior Work: 

- For a list of CIPPIC's past work, see [here](https://preview.cippic.ca/articles/cippic-s-past-work). 

#### Ongoing Matters:

- _CIPPIC v Sahni_ (more [here](https://www.cippic.ca/articles/cippic-v-sahni-ai-s-role-in-copyright-law)) (seeking to expunge registration of copyright by an AI “author”). 
- _Blacklock’s Reporter v Attorney General of Canada_ (more [here](https://www.cippic.ca/articles/fair-dealing-and-digital-locks-cippic-s-intervention-at-the-federal-court-of-appeal), [here](https://www.cippic.ca/articles/cippic-granted-leave-to-intervene-in-blacklock-s-reporter-appeal), [here](https://www.cippic.ca/articles/federal-court-rules-that-digital-locks-do-not-restrain-fair-dealing-using-a-validly-obtained-password), and [here](https://www.cippic.ca/articles/cippic-discusses-digital-locks-case-on-michael-geist-s-law-bytes-podcast)) (balancing copyright enforcement with user rights).
- _Voltage_ Matters (more [here](https://www.cippic.ca/articles/supreme-court-refuses-to-hear-voltage-v-doe2), [here](https://www.cippic.ca/articles/cippic-granted-leave-to-be-added-as-a-party-by-the-supreme-court), [here](https://www.cippic.ca/articles/cippic-files-respondent-factum-voltage-v-doe-scc-proceeding), [here](https://www.cippic.ca/articles/appeals-court-affirms-denial-of-default-judgement-in-file-sharing-case), [here](https://www.cippic.ca/articles/voltage-reverse-class-lawsuit-denied-certification-again), [here](https://www.cippic.ca/articles/voltage-reverse-class-proceeding-denied), [here](https://www.cippic.ca/articles/federal-court-of-canada-sets-rogers-reasonable-costs-in-norwich-orders), and [here](https://www.cippic.ca/articles/supreme-court-of-canada-decision-in-rogers-v-voltage-clarifies-role-of-is-ps-in-copyright-infringement)) (protecting Internet subscriber privacy rights). 
- Various CRTC Matters (more [here](https://www.cippic.ca/articles/championing-consumer-rights)) (making information for Internet and mobile consumers easier to access and understand).
- _Malone v Library and Archives Canada_ (preventing the destruction of copies of records released under the _Access to Information Act_). 

#### Recent Commentary / Blogs:

- February 22, 2025 - Student Tina Derak on [Upcycling in Trademark Infringement](https://preview.cippic.ca/articles/fashion-upcycling-and-trademark-infringement). 
- February 14, 2025 – Articling Student Gareth Spanglett on [AI and Copyright](https://www.cippic.ca/articles/the-ai-copyright-conundrum).  
- January 15, 2025 – Director Matt Malone on [TikTok in the United States](https://www.theglobeandmail.com/business/commentary/article-while-the-us-supreme-court-deliberates-a-tiktok-ban-canadas-own-tiktok/). 

#### Mentorship:

- CIPPIC is working with a talented group of law students this 2025 winter term. 
- CIPPIC is helping coach law students for three moots during this 2025 winter term (building on some [record-setting success](https://www.uottawa.ca/research-innovation/news-all/law-students-repeat-winner-oxford-international-intellectual-property-moot)).

#### Ongoing Initiatives:

- CIPPIC’s [Open Information Initiative](https://www.cippic.ca/articles/the-launch-of-cippic-s-open-information-initiative).  

#### New Research:

- Secret. More soon 😊 

#### Questions:

- Any questions? Contact admin@cippic.ca.  


small_Working.png

thumbnail_Working.png

Working.png