By Steve Oetting
Is Artificial Intelligence a misnomer? Or an oxymoron? Or could it be an oxymoronic misnomer?
On October 1, I was fortunate to be a participant in a webinar about AI. But this informative seminar was not about the validity of AI, rather it examined how AI affects creators and publishers in today’s convoluted legal arenas.
The event was moderated by Kate Edward of Access Copyright, a leader in content licensing solutions for Canada. Ms. Edward welcomed the over 500 participants before explaining a bit about her company, what their purpose was and why they felt it was important to bring this information to the forefront. She offered up two pertinent survey questions to the group.
The first question was, “How familiar are you with Generative AI Tools, such as ChatGPT, Gemini, Claude, et cetera?” Twelve percent responded “Very familiar”, 51% “Somewhat familiar”, 18% “Somewhat unfamiliar” and 19% “Not familiar at all”.
The second question was, “What best describes your feelings about Generative AI?”, to which 49% responded “Deeply concerned”, 24% “Somewhat concerned”, 4% “Neutral”, 16% “Cautiously intrigued”, and 9% “Very intrigued”.
Suffice it to say that many of the people attending were deeply concerned about Generative AI. This response made sense since this audience was invited to a seminar that would explore AI from the perspective of copyright infringement and litigation. Ms. Edward then introduced Erin Finlay, a partner of Stohn Hay Cafazzo Heim Finlay LLP, legal experts in this area of Canadian law.
Ms. Finlay opened her presentation with a discussion about copyrighting and what protections it provides. There are clear standards established by thousands of copyright infringement cases that define precisely how copyrighted material must be duplicated in order to be qualified as a copyright infringement. These cases typically refer to the reproduction of exact wording for a reasonable volume of the material (ie. a minimum number of words). Cases where the replication of material was exact enough and substantial enough would most often favor the position of the complainant.
However, Generative AI does not duplicate exact wording, rather it scrapes (collects) its data in chunks, or phrases, that are not exact enough to qualify as copyright infringement. In fact, no case of copyright infringement involving Generative AI has ever been won in Canada. It is simply impossible to compare AI-generated material to the original material in order to define it according to traditional copyright infringement standards. In fact, many legal firms, such as Ms. Finlay’s, are hesitant to advise clients to proceed with litigation against AI-generated material as copyright infringement.
But this is changing. Very recently, a federal judge in California preliminarily approved a landmark $1.5 billion settlement of a copyright class action brought by a group of authors against artificial intelligence company Anthropic, the developers of Claudia. The case was not based on the duplication of the copyrighted material, but on the fact that Anthropic fed copyrighted material to its AI engine that was pirated, ie. not properly purchased, and for which no written approval had been provided by the authors.
While AI companies continue to dispute the accusations of copyright infringement, their tools are being constantly improved with a particular mandate to avoid detection of duplication and the identification of originating material. Simultaneously, tools are also evolving that can better detect the use of copyrighted text in AI-generated material. But in Ms. Finlay’s expert opinion, the chances of anyone winning a case of copyright infringement against AI-generated material in Canada today are less than slim, they are non-existent, as evidenced by the fact that no such case has ever been won by a complainant.
“What does this mean to a group of writers such as the Manitoba Writers’ Guild? I believe there are two salient conclusions we can draw from Ms. Finlay’s analysis. The first is that it is likely impossible for us to accurately assess whether a piece of written material is AI-generated at this time. The second is that because this arena is perpetually changing, we must continue to remain informed and educated. Only if we stay current with AI do we stand even a chance of managing this foreboding challenge coherently and successfully in the future.