AI and Copyright

I was 11 years old when, in the summer of 1983, I signed up for a computer programming course at a local community centre in Winnipeg, where I lived at the time. I’ll never forget the thrill I felt at sitting in front of a Commodore 64 keyboard, with one of those green phosphor monochrome monitors and a small cassette tape player for a data drive, and learning how to write a simple program. The computer language was BASIC. The main prompt of my program was a text string that would pop up saying “How are you today?”. If the user responded with ‘good’ the program would congratulate them and suggest they reward themselves with an ice cream cone. If the response was bad, the program would offer reassurances that things would probably get better tomorrow. If the response was something else, the program would say it didn’t understand and ask the user to try again.

It was a ridiculously simple interaction—yet, even though I created the thing, the computer’s apparent ability to assess and respond fascinated me. Sure, I wrote the initial instructions, but still it occurred to me that the computer was doing something else. In between responses, it waited. It would wait as long as needed until someone entered the next text string. Technically, there is no magic here, just a machine that can’t do anything until its next input. But if you squinted just right you could almost see a little person in there listening, considering, and waiting for me.

It instilled in me a question that now seems to reverberate around the world: can these things actually think?

While we’re not analyzing machine intelligence here (spoiler alert: the answer is no, they can’t think), it seems that a similar question (and fascination) now motivates what could be the next technological revolution. There are few industries now that aren’t proud to show off some new shiny implementation of so-called ‘artificially intelligent’ processing.

Yet, such technology, while lightyears beyond my little BASIC program, is not a panacea. It carries a number of ethical, environmental, interpersonal, economic, and legal problems. The legal problems are plentiful and include (in addition to others) issues around privacy, misinformation, fraud, environmental liability, workplace relations, negligence, medical liability, patent eligibility, administration of criminal law, immigration eligibility, and copyright controls.

As of this writing, the AI boom remains in its infancy. Many of the legal issues have yet to be litigated and work their way through court systems and legislatures. No doubt, there will be new laws implemented in response to flurries of lawsuits. There will need to be.

But for now, we can still take a principled look at the legal landscape and examine at least one area where AI considerations are especially important: the law of copyrights.

What is AI?

To start, let’s describe what we mean by ‘artificial intelligence’ (“AI”). AI can be defined as: “a field of science concerned with building computers and machines that can reason, learn, and act in such a way that would normally require human intelligence or that involves data whose scale exceeds what humans can analyze.[1]” It works through exposure to vast amounts of data, identifying “patterns and relationships that humans may miss” or by generating new data through predictive reasoning.

A subcategory of AI is so-called ‘generative AI’—in which a user provides a prompt describing the desired outcome and the AI computers create a new work based on the prompt. The user can then save the work for later use or reject it and have the AI try again until the user is happy with the result.

On the backend, generative AI is processed in gigantic, networked data centres, distributing processing between computer blocks and even across borders. The scope of this processing is truly vast. It is estimated that by 2027 the global demand for water used in cooling AI computers could be equivalent to half that used in the entire United Kingdom[2]. Indeed, Google and Microsoft are investing in nuclear reactors just to power their AI engines.

Critical to our analysis is that AI data processing centres can be located anywhere in the world. For example, according to the Observer, “OpenAI has not disclosed the exact location of its data centers, but has partnered with Microsoft in the past to distribute servers worldwide on Amazon’s Azure platform.[3]”

Application to Copyrights

Copyright in Canada is defined as: “the sole right to produce or reproduce the work or any substantial part thereof in any material form whatever, to perform the work or any substantial part thereof in public or, if the work is unpublished, to publish the work or any substantial part thereof[4]. . .”

To qualify for copyright protection, the particular work must be a “literary, dramatic, musical [or] artistic work. . .[5]” Also, the work must be original[6] and the author must be “at the date of the making of the work, a citizen or subject of, or a person ordinarily resident in, a treaty country.[7]”

Application to AI

We can see that in Canada for copyright protection to subsist in any work, that work must have an author and the author must be a ‘citizen or subject of, or a person ordinarily resident’ in a treaty country.

In Canada, a citizen is defined (albeit circularly) as a “Canadian citizen”[8].  Section 3(9) of that act adds some texture to the definition by listing criteria for being a citizen. Each criterion requires that a citizen also must be a ‘person’.

The law of ‘personhood’ has a deep background and a bit of a dark history: it has been used (among others) to exclude members of society from political or economic participation based on definitions of personhood. As noted by Kathleen Lahey[9]:

“Legal personhood” or “civil capacity” has been used to regulate participation in the state or civil society for almost as long as written law codes have been used to regulate human conduct — some 4500 years.  Without civil status (“legal personhood” or “legal personality”), people cannot access the legal process, but merely exist in the shadows of the law, subject to the whims of anyone who has more personal power, or who can invoke the power of the state to back them up.[10]

As such, historically courts have tended to limit definitions of personhood rather than expand them in order to justify the desired policy outcome of the case. More recently, though, courts have taken a wider view, holding that in Canada a ‘person’ is “a human being or a natural person as well as an artificial person such as a corporation. The primary sense of the word is a natural person; the secondary sense, an artificial person such as a corporation.[11]”

From this perspective, clusters of computers are not, and cannot be, persons within the meaning of copyright laws. They are not even a single thing. Also, they have none of the social or political needs of human beings and none of the economic utility of well-regulated corporations. They are merely mechanical information processing tools.

The fact that AI cannot be a ‘person’ excludes its creations from copyright protection in Canada. But let’s go a step further because in Canada, copyright protection also requires an ‘author’.

Is AI an author?

Authorship is something murkier than personhood or residence. It relates to the human experience and how we live it. Authorship speaks to the broader value of a work and the reasoning for that value – extending from a person’s insights into their lived experiences, combined with the idea that creative works form a kind of language speaking to these insights. We value creative works because they present insights and experiences that we cannot get from other means. If you speak with artists and creative folks, you will hear a great deal about ‘discovering’ and ‘learning’ in the creative process.

So in a very real sense, authorship is about participating in life —which AI cannot do. It seems the best it can hope to do is to calculate what comes next in a series of processing streams and create something that satisfies output criteria.

Craig and Kerr[12] addressed this pretty well, concluding that:

To live means to participate in dialogue: to ask questions, to heed, to respond, to agree, and so forth,” wrote Bakhtin[13]. We might equally say that to participate in dialogue of this nature means to be alive — something that artifacts, by definition, cannot be. If the great value of authorship to society lies in encouraging the human creative capacity “to resist and transform existing patterns,[14]” this value is nowhere to be found in AI processes that merely identify existing patterns, reinforce them, and replicate them.

Indeed. Generative AI seems like a juggernaut, but it’s important to remember that it has limitations. It is not a do-all, nor, I suspect, is it the artist-killer that some fear.

A real live artist, after all, with years of training, experiences, and good-old hard knocks, will always be able to deliver enforceable copyrights[15].

The take-away from this is that users of Generative AI need to be aware that their images will not qualify for copyright protection without substantial creative contributions from real people. Also, it’s worth noting that without legislative clarity and business vigilance, GenAI could severely dilute the well of commercially controllable intellectual property.

---

[1] Google Cloud: https://cloud.google.com/learn/what-is-artificial-intelligence

[2] Crawford, Kate “Generative AI’s Environmental Costs are Soaring — and Mostly Secret”, Nature: https://www.nature.com/articles/d41586-024-00478-x

[3] Montie, Eve, “Elon Musk’s xAI Announces New Data Center: Where Do A.I. Giants Store Their Data?” The Observer: https://observer.com/2024/06/elon-musk-xai-big-tech-data-center-location/

[4] Copyright Act (R.S.C., 1985, c. C-42), s. 3(1)

[5] Copyright Act (R.S.C., 1985, c. C-42), s. 5(1)

[6] Copyright Act (R.S.C., 1985, c. C-42), s. 5(1)

[7] Copyright Act (R.S.C., 1985, c. C-42), s. 5(1)(a). A ‘treaty country’ is “a Berne Convention country, UCC country, WCT country or WTO Member”

[8] Citizenship Act (R.S.C., 1985, c. C-29) s.2(1)

[9] Faculty of Law, Queen’s University, Kingston, Ontario

[10] Lahey, Kathleen, Legal “Persons” and the Charter of Rights: Gender, Race, and Sexuality in Canada (1998, CanLIIDocs 80)

[11] Kennedy v Canada (Customs and Revenue Agency), [2000] OJ No 3313

[12] Carys Craig & Ian Kerr, “Death of the AI Author” (2021) 52 Ottawa L Rev. 31 – 86 

[13] Gary Saul Morson & Caryl Emerson, Mikhail Bakhtin: Creation of a Prosaics (Stanford: Stanford University Press, 1990) at 60.

[14] See Jennifer Nedelsky, Law’s Relations: A Relational Theory of Self, Autonomy, and Law (New York: Oxford University Press, 2011).

[15] Subject, of course, to citizenship and other requirements of relevant copyright laws.