Canadian artist wants Anthropic AI lawsuit corrected
Tim Boucher objects to the mischaracterization of his work in authors' copyright claim
Interview Tim Boucher, a Canadian artist, author, and AI activist, sent a letter to the San Francisco judge overseeing an authors' lawsuit against AI firm Anthropic to object to the way the legal filing characterizes his work.
The complaint, which accuses Anthropic of violating authors' copyrights, says that the company's Claude model "has been used to generate cheap book content." As an example of the economic impact that AI-generated work has had on authors, it cites a Newsweek report about Boucher using Claude to write 97 books in less than a year.
In his letter [PDF] to Judge William Alsup, Boucher says that passage "mischaracterize[s] my work in a manner that has led to reputational harm, including causing a major media outlet to refer to me incorrectly as a 'fraudster.'"
The letter – a request rather than a defamation lawsuit at this point – scolds the plaintiffs' attorneys (cc'd) for the unwarranted use of his name to support their claims against Anthropic. And it asks for a revision of the complaint.
Acknowledging that the proliferation of AI-generated junk books on Amazon is a real problem, Boucher contends that his work should not be cited as an example of deceptive, damaging AI content.
"I do not copy or rip off the books of others," he wrote. "The contents of my books come from my imagination and I use AI tools to realize that vision. I do not falsely attribute my books to other authors. I do not sell my books on Amazon."
What's more, Boucher argues that since he has never paid Anthropic for AI services, his work should not be used to support allegations that Anthropic has profited from copyrighted work.
The lack of care exhibited by the plaintiffs' attorneys, he contends, "was unnecessary, unjust, and cruel. I ask that they correct the record and show greater consideration in the future for the real human beings affected by their litigation tactics."
The Register asked two of the attorneys involved in the complaint whether they'd care to comment. We've not heard back.
Following The Register's reporting on the lawsuit, Boucher reached out to correct the record about his work. He agreed to be interviewed via email, prior to sending his letter.
The Register: I understand that you object to the way the attorneys suing Anthropic have described your work in their complaint. Can you elaborate on your concerns and how you intend to respond?
Boucher: The complaint identifies some real problems, particularly on Amazon, where unscrupulous users of generative AI are creating copycat or rip-off works, and falsely attributing them to popular authors.
However, it then goes on to wrongly use my work as an illustration of this trend. I don't engage in any of those practices. I don't even sell my books on Amazon. What's in my books comes from my imagination, and I use AI tools to realize that vision, just like I might in other cases use a paintbrush and canvas, or linoleum block cuts to do the same.
I am sending a notice to the lawyers involved in the suit about the real reputational harm their mischaracterization of my work has caused me, and asking them to amend their filing. If they do not do so, I will be sending a request to Judge Alsup to be allowed to enter a statement into the public docket for the case, correcting the inaccuracies contained in the filing. If necessary, I will also pursue the matter in the Court of Quebec, where I reside.
The Register: Do you have a view on whether AI models violate copyright law in Canada or other jurisdictions, either for training or inference?
Boucher: I'm not an expert on copyright law, but I did submit a statement as an artist using generative AI tools to both the US Copyright Office's public consultation on generative AI, and the one run by the Canadian Intellectual Property Office last year. I participated in that because it's important that artists of all kinds (whether or not they choose to use AI) have a voice, since they ultimately drive a lot of innovation, and are rarely recognized for doing so.
Here's a link [PDF] to my individual submission, reproduced on Berkeley Law's website.
To paraphrase from that document, and as stated in the ad hoc Artists Using Generative AI group submission to Copyright Office:
Copyright law should continue to leave room for people to study and analyze existing works in order to craft new ones, including through the use of automated means like those used to create AI models.
Copyright law is complicated, and varies substantially between countries. Again, I'm an artist and not a lawyer, but based on my research under US law, most uses of copyrighted materials for training purposes seem to qualify as Fair Use, and not infringing. The purpose of including items in AI data sets is not to copy, store them for retrieval, or reproduce them. Its aim is to analyze, measure, and compare their properties in aggregate in order to transformatively create new works which are not merely derivative of works in the training data but entirely new.
My position on the copyrightability of outputs from generative AI systems is that they are transformative, and that they ought to be able to be registered with the US Copyright Office. My books as a series are in fact registered with CIPO, which does not review or assess copyright submissions like the USCO does. I could be wrong, but I believe that under the Berne Convention, this automatically grants me copyright protections as a creator in the US, which would not otherwise be available were I to create the works while physically present there, due to the USCO's overly restrictive rulings on [Gen]AI. There needs to be more scholarly thought around how best to harmonize these different regimes in a way that makes more sense than it does now.
As to Canada, I like the position taken by the Canadian Bar Association here [PDF] that characterizes the use of generative AI as being similar to what a cinematographer might do when they compose and arrange film or video works, and that this should be adapted for AI-generated or assisted works.
The complexity and collaborative nature of creating a cinematographic work compares well with the challenges posed by AI created works. For cinematographic works, the Canadian Copyright Act states that copyright subsists in the work's 'maker' – which can even be a corporation. In relation to cinematographic works, the Act defines a maker as 'the person by whom the arrangements necessary for the making of the work are undertaken.' Interestingly, the United Kingdom Copyright, Design, and Patent Act deems the 'person by whom the arrangements necessary for the creation of the work are undertaken' to be the author of any computer-generated work.
- Salesforce mulls charging per AI chat as investors sweat over fewer seats
- Fintech outfit Klarna swaps humans for AI by not replacing departing workers
- Brit teachers are getting AI sidekicks to help with marking and lesson plans
- Facebook whistleblower calls for transparency in social media, AI
The Register: What about the concept of moral rights, which in the US is limited to the right to attribution and the right to integrity, and applies to visual works (17 U.S.C. § 106A), but extends to all works in Canada? Do you see an issue with AI as it applies to moral rights?
Boucher: In France, where my books are being translated and printed by Typophilia, my understanding is that moral rights are an even stronger concept than they are in the US or Canada. Unlike economic rights, which can be sold or transferred, moral rights of French creators over their works are perpetual, inheritable, and cannot be transferred or sold. I think the French centering of this subject on the rights of the author is really interesting in that it is so different from how copyright law works in the US, and that it can be so empowering for creators to have some measure of downstream control – even where it might conflict with economic rights owned by a third party. But I think it's still not a settled matter in France as to how moral rights exactly should be adapted into these training scenarios for AI models, where billions of works are thrown into a blender, and become a kind of statistical soup that is drawn from and mixed through inference with data from user-defined generative AI prompts.
There is a piece by technologist Jaron Lanier, that I substantially agree with, where he advocates for there being something like an "attribution layer" in generative AI systems:
There is no reason to hide which artists were the primary sources when a program synthesizes new art. Indeed, why can't people become proud, recognized, and wealthy by becoming ever-better providers of examples to make AI programs work better?
I think that's an idea that is worthy of exploration, though I recognize that maybe in the present configuration of these technologies, it might be difficult or perhaps impossible to pull off. My understanding of generative AI technology is that its outputs are transformative of the billions of points of data required to train these models. They are not generally reproducing any specific work(s) in part or in entirety, nor are they based on any small given subset of works, but on broad statistical patterns distributed across the entire training set. That said, we should be able to design the technologies to work the way we want them to work, rather than be forced to accept them working in ways we don't all agree on. To say that it can't work a certain way because it does not presently is a form of technological determinism I strongly disagree with.
The Register: Is the output of AI models useful or valuable to you as an artist? If so, in what way?
Boucher: AI has made me a vastly better writer. I've been writing for a few decades now, personally and sometimes professionally. But there are certain things I've always fallen short in, certain forms of structured writing and logical flow of arguments especially which have always eluded me. LLMs tend to excel at this kind of writing, even if their outputs can sometimes tend toward the vanilla. So the ability to have this tool, this writing partner, to bounce my ideas off of, and who can rapidly produce semi-usable results has been incredible. It's not strictly a question of enhancing productivity or volume of work that I can create (though it's that too), but this interrogative way of working has rubbed off on me, and the AI tools have taught me how to actually think more logically and clearly about problems, and then to more plainly organize those thoughts and communicate them with others. The same process is largely at play with using image generators as well, where it has helped develop my eye as an artist and artistic director, where I can look at a visual piece – whether a painting I did, or a [Gen]AI project – and more easily evaluate, does this work for the effect that I want? And then to be able to rapidly iterate on the results until it matches my vision.
The Register: Would you be willing to share any information about how much of your income these days comes from the distribution and sale of AI-augmented or generated work?
Boucher: I appreciate the interest that people have in specific financials for sales of AI books, as this information is still hard to come by. I made the mistake early on in my Newsweek article of going into some details about that in the interest of transparency, and what I saw was that it provoked two possible sets of angry reactions: 1) this guy is making too much money off AI, and 2) this guy is not actually making that much money off AI. Neither of those responses really drive the conversation forward in a meaningful way, I think, nor do they get at the more complex issues around AI and copyright, and what role we'd like to have AI play in relation to human creativity. So I'd just as soon avoid rehashing those arguments again.
I will say that my books continue to sell and get media coverage without any advertising. And that a significant portion of my buyers become repeat buyers, with many, many examples of readers coming back to buy ten, 20, or in one case even nearly 50 different titles. That tells me people are excited about the format and contents of my books, and that's much more important than specific sales numbers, which can only go up with time.
The Register: Does it concern you that using AI to generate books floods the market and dilutes economic opportunity for authors writing without AI assistance? Or do you see your work as separate from the overall media/attention ecosystem?
Boucher: I think it's very concerning when people do actually create knock-off works which misuse the names of authors. The complaint is correct there. Those practices absolutely flood the market and dilute the economic opportunity for those authors. But that's not what I'm doing at all.
My books are something else entirely. My books are marketed as AI-assisted books and include disclaimers prominently stating that fact. The print versions in France even have stickers affixed to the covers alerting readers to this fact. To me it's a selling point, not a drawback. Not everyone will be into it, but being transparent and up front about accurate labeling is always good practice, and should be adopted throughout the industry.
What I think is that, as these technologies improve, many readers might find they enjoy AI books as much or more than those produced by human authors. Some of the AI "minds," for example, in Iain M Banks' Culture books become accomplished artists. There's a really compelling paradigm shift that generative AI invites us into as readers and viewers, where the act of creating and consuming become sort of fused. The human author becomes kind of the "first reader," who then helps point the way forward for other readers. Or it might be that works created this way will be privately produced for an audience of one, and that's just fine too. I support any and all permutations of how that could work for artists and audiences, including for those who choose not to use these tools at all, and to avoid any work created using AI. There will always be markets for both types of creation.
The Register: Do you read/view/listen to other AI-generated content when you're reading/viewing/listening for pleasure? Or do you prefer human-authored works?
Boucher: I've read a few AI-generated books, but they weren't memorable for me because they were not doing anything especially new or interesting with the form. That's a big innovation that using AI has been able to afford me: I can now play with these larger meta-forms through my work, because a certain amount of boots-on-the-ground labor has been liberated. I think what we're seeing now is genuinely the birth of a new art movement and would love to see people who are working in this space really push the limits of the technology and our conventional understanding of what "books" and even "art" are. But we still need to infuse ourselves, our personalities, our thoughts, our emotions, our humanity into these works to make them really engaging and exciting.
The Register: Did you use AI to assist with any of these answers?
Boucher: Naturally, I tried. I took the questions, along with other prior pieces of writing I had in this space, and tried to get ChatGPT to produce something usable. Sometimes that works, but often it doesn't. In this case, the answers fell very short and were extremely flat and weird and boring. So I did go through a bit with it just to think out loud about certain points, but didn't end up using any actual text generated by AI to answer these questions.
The Register: Any further thoughts you'd want to share?
Boucher: Already covered a ton of ground here, so I don't think there's anything really missing. Really appreciate you taking the time to hear me out. I'd like to help reduce the stigma for artists who want to explore and create in this new emerging media that I think of as a "hypercanvas," where instead of individual brushstrokes, each image, or each query in an LLM becomes the "brushstrokes" in these higher dimensional works. I've taken a lot of flak publicly for being up front about what I'm doing, but I'm a firm believer that these conversations need to happen publicly about what we want from AI (or from not using AI), and how best we can get there. So, if what I'm doing can open the door for other artists to not be scared to experiment with these incredible and incredibly imperfect new technologies, and to not be constantly insulted, threatened, and mischaracterized, well then, I think I will have achieved something in the end. ®