Canadian Institute for Advanced Research

Les informations de ce site Web à été fournie par des sources externes. Le gouvernement du Canada n'assume aucune responsabilité concernant la précision, l'actualité ou la fiabilité des informations fournies par les sources externes. Les utilisateurs qui désirent employer cette information devraient consulter directement la source des informations. Le contenu fournit par les sources externes n'est pas assujetti aux exigences sur les langues officielles et la protection des renseignements personnels.

Consultation on a Modern Copyright Framework for Articial Intelligence and the Internet of Things: CIFAR Response


The intent of this submission is to provide an overview of current practices and attitudes within the articial intelligence (AI) research community regarding copyright. Providing legislative or policy recommendations is beyond the scope of the paper.

CIFAR conducted a number of interviews with AI researchers to better understand the current state of leading edge AI practice as it relates to the Copyright Act. Interviews focused on text and data mining and the AI-driven or AI-assisted generation of new works.

CIFAR’s Relationship to the AI Community

CIFAR fellows have been instrumental in the development of modern AI. In the early 2000s, Geoffrey Hinton approached CIFAR with an idea. He had been a member of CIFAR’s rst program, Articial Intelligence, Robotics & Society (founded in 1983) and he had become convinced of the power of neural networks and their potential for deep learning in machines. By early 2004, Hinton was leading CIFAR’s Neural Computation & Adaptive Perception program (NCAP).

Its members included Yoshua Bengio and Yann LeCun, among other neuroscientists, computer scientists, biologists, electrical engineers, physicists, and psychologists. Today, the three are widely acknowledged as the pioneers of deep learning.

for leading-edge AI research. CIFAR also leads Canada’s national AI research and innovation strategy, the Pan-Canadian AI Strategy, working closely with Canada’s three national AI Institutes (Amii, Mila and the Vector Institute) and our network of 100+ Canada CIFAR AI Chairs across the country. Through these, CIFAR continues to build scientic capacity in AI, while encouraging its ethical application.


Text and Data Mining

Researchers were clear that access to data is a critical input to AI research, and that improved access to data sources for research is a benet. That said, the existing copyright regime was not typically viewed as an impediment to academic research. There were four examples of methods highlighted to access data for text and data mining purposes.

First, a signicant portion of machine learning is conducted on standardized datasets (such as CIFAR-10 or MNIST, in the case of image recognition) to permit comparison of results. These generally have permissive licenses, if any are specied at all, are widely available and used, and their copyright status is not an area of concern.

Second, the use of publicly viewable copyrighted works was typically acquired through a non-nancial licensing process with a rights holder. Twitter, for example, is a common source of text data, and access to its API requires agreement and adherence to a licensing agreement.

Third, additional licensing agreements were entered into to provide access to text or image data that was not available to the general public. These agreements typically had considerations beyond copyright, most notably privacy.

Finally, researchers collected data from open and publicly available sources, such as scraping websites. A number of researchers did not appear to acknowledge that these publicly available works are, in most cases, copyrighted. Though the use of scraped works for research purposes could be permitted as fair dealing, this did not arise during interviews.

AI-Generated Works

In the case of AI-driven or AI-assisted generation of new works of text or images, the current copyright framework was also not considered to be a barrier or benet to academic work. In particular, researchers did not view a lack of copyright protection for these works as a disincentive to their research. While beyond the scope of this consultation, trade secrets is typically a much more important tool for IP protection in AI research than copyright or patents.

Additionally, researchers did not consider themselves to have an authorial role in the generation of synthetic works; however, there were indications that should the quality and perceived merits of these works improve, researchers may change their views regarding their role as authors.


To ensure Canada’s continued competitiveness in AI research, along with associated benets to Canadians’ social and economic lives, it is critical that access to data for research purposes be responsibly expanded wherever possible. It is also critical to recognize that Canada’s copyright framework exists within a global context of competition for talent and that, where possible, Canada should pursue reforms through multilateral actions and treaties.

While copyright was not viewed as a signicant barrier (nor incentive) to AI research, there may nonetheless be opportunities to unlock data assets for research that continues to ensure the fair treatment of rightsholders and the ongoing incentive to create and publish new creative works.


CIFAR is a global research organization that convenes extraordinary minds to address the most important questions facing science and humanity.

By supporting long-term interdisciplinary collaboration, CIFAR provides researchers with an unparalleled environment of trust, transparency and knowledge sharing. Our time-tested model inspires new directions of inquiry, accelerates discovery and yields breakthroughs across borders and academic disciplines. Through knowledge mobilization, we are catalysts for change in industry, government and society. CIFAR’s community of fellows includes 19 Nobel laureates and more than 400 researchers from 22 countries. In 2017, the Government of Canada appointed CIFAR to develop and lead the Pan-Canadian Articial Intelligence Strategy, the world's rst national AI strategy.

CIFAR is supported by the governments of Canada, Alberta and Quebec, as well as foundations, individuals, corporations and Canadian and international partner organizations.