The information on this Web page has been provided by external sources. The Government of Canada is not responsible for the accuracy, reliability or currency of the information supplied by external sources. Users wishing to rely upon this information should consult directly with the source of the information. Content provided by external sources is not subject to official languages and privacy requirements.
Submitted To: Innovation, Science and Economic Development Canada and Canadian Heritage
Submitted By: Aaron Shull, Managing Director and General Counsel, and Bob Fay, Managing Director Digital Economy Research
Re: Consultation on a Modern Copyright Framework for AI and the Internet of Things
Date: September 16, 2021
Thank-you for the opportunity to respond to this consultation on a Modern Copyright Framework for AI and the Internet of Things.
Please find comments below for your consideration that are based largely on a discussion that CIGI convened with a group of seven multidisciplinary expertsFootnote 1 to review this consultative document. Our main recommendation is that we should avoid rushing modifications to the current Copyright Act.
The Copyright Act is a statute that has a long history, and it is structured in a particular way that is linked to international agreements and international systems. Any premature or unnecessary amendments to the Act may be obsolete and irrelevant 5-10 years from now - and, worse still, have unintended harmful consequences in the meantime. Indeed, artificial intelligence (AI) continues to evolve at a rapid pace, and we have yet to fully grasp its potential. In fact, we do not even yet have a common definition of AI making it difficult to ascertain how to fit it into current IP arrangements. Often, we rush to add to our current legislation every time there is a new technology. Although it may seem to be counterintuitive, it can sometimes be better to wait and see how the technology unfolds. The development of the VCR and the movie rental business in the 1980’s is a classic example of when no legislative action, despite hue and cry to the contrary related to copyright infringement and “pirate” copies, was the best action.
More generally, there has been a view that new technologies inevitably harm old ones, and that legislation is required to address this when the marketplace may solve the issue. Within the context of text and data mining (“TDM”), it is too early to legislate for specifics, and we run the risk of putting in exceptions that are unnecessary, redundant, and potentially dangerous - at best. We need to gather more evidence on how AI is being used and deployed and how current intellectual property arrangements are affecting the uses of AI. For example, training AI may not require the reproduction and storage of works that could infringe on copyright. The most that should be done about TDM is to include the words “such as” in s. 29 of the Copyright Act – an amendment that is long overdue for many other valid reasons as well.
- The consultation document does not integrate the theoretical foundation of copyright policy
Nowhere in the document does the discussion focus on the theoretical foundation of copyright policy. Rather the document goes through each of the identified issues (which is a good start but there are certainly more) and presents solutions. This is especially problematic when it comes to dealing with AI-generated work without human intervention (for example), as it questions what the underlying interests are; whether there is a clear incentive; where the rewards are; and what we are looking at in terms of underlying policy. Are we just assuming authorship from AI that can be authorized under the Copyright Act? What is the public policy objective here? What are we trying to achieve? It should be recalled that the 1709 Statute of Anne – which is the basis of copyright law as we know it – was all about the protection of “authors.” Authors are humans.
- Artificial intelligence relies on a tremendous amount of data to train algorithms - which has raised questions about whether or not the permission of rights holders is needed
Google translate is remarkable at translating text. It was developed by relying heavily on AI that scanned millions of documents indexed on the web. However, no one would suggest that Google should own copyright of the resulting translation. Some legal scholars have contended that Google and other companies using AI might already be covered by the fair-dealing provision in the Copyright Act, which allows for the use of copyrighted materials for the purpose of research or study. The Supreme Court of Canada also addressed the fair-dealing exemption in its ruling in CCH Canadian Ltd. v. Law Society of Upper Canada, and recently in York University v. Access Copyright; however, the questions surrounding who owns AI-generated works are complicated.Footnote 2
- There needs to be a better way to widen the perception of ‘fair dealing’ without adding additional terms such as ‘informational analysis’
Informational analysis appears to be a very broad term that should be avoided. It could significantly dilute the value of data for data owners. We need to balance the rights of data owners with other data rights, particularly privacy, and contractual rights around data. The data owners are different than the data subjects, which adds another layer of complexity to what is considered ‘informational analysis’ and what is permitted under the Copyright Act. Once again, all that is prudent or necessary at this time is to add the words “such as” to s. 29.
- There is a need to anchor ownership and authorship of AI generated work with the notion of originality and having a human involved in their development
Music production software has been around for a long time (for example), and has been used to develop new works, but we still need to anchor these works with the notion of originality and having a human involved in their development. The courts have been very generous in considering what is deemed skill and judgement to create originality. Particularly, this was seen in Geophysical Service Incorporated v Encana Corporation, 2016 ABQB 230, where the Alberta Court of the Queen's Bench had the opportunity to assess both claims to copyright in seismic data and the power of certain regulatory boards to release copies of such data to the public.
There is a need to balance rights surrounding data with this notion of getting a data mining exception, while also potentially broadening what copyright applies to if we are going to expand copyright to apply to AI generated works where there's no human in the loop. Once again, there is no justification for elimination of the requirement of human authorship in works that involve the use of AI, which should be regarded as no more than a very powerful tool to enable and enhance the human element of creativity. At the end of the day, a work generated autonomously by AI with no or insufficient human creative input should not be protected by copyright.
- There is a need to establish an auxiliary right to data
There is substantial benefit to feeding an immense amount of data into AI; however, there is not a lot of benefit (for traditional copyright reasons) to prevent people who are conducting text and data mining, from accessing copyright protected works. We need a framework that is independent of the Copyright Act and comes at it from a new perspective. A possible solution could be to establish an auxiliary right to data -which could help to create a marketplace for high-quality information. This does not need to be a full copyright, right; rather a supplementary right to access that would allow its use in AI but maintain economic rights under copyright. For the foreseeable future, all that is prudent or necessary is the addition of the words “such as” in s. 29.
- We are living in a world in which we assume that all we need to do is keep on increasing rights, appropriation and capturing creative expression; however, the structure around intellectual property rights and copyright (in particular) is about limitations
We need to find a solution to AI driven or AI authored works within a body of law. There is generally no justification under copyright law to ascribe authorship to a completely AI created work. There are different rationales and policy reasons as to why we might look at a different option that would encourage people to continue to develop AI, enable people to work with what is developed, and build new knowledge. While copyright may be inadequate, or the wrong model, we cannot just keep on adding pieces without justifying the underlying rationale. We may someday need a more principled approach to policy surrounding AI, but one driven by a full understanding of incentive, encouragement, and innovation. Any premature legislation could have significant unintended consequences.
- Some of the issues addressed in the consultative document are outside of the realm of copyright
While implementing a ‘right to repair’ could be a useful amendment to the Copyright Act (even though it is typically an anti-trust issue, it could, however, be advisable for other reasons. It would clarify for greater certainty that circumvention of a technical protection measure (“TPM) is not an infringement of copyright if done for the purpose of fair dealing pursuant to s. 29.
In the end, copyright reform should be driven by principle. It has always been about providing certain protections, in order to encourage new and novel creations. We should be able to concretely answer two fundamental and interrelated questions: What is it that we are trying to encourage by adding this protection? And how does this further encouragement lead to a stronger, more vibrant, and more prosperous nation.