Artificial Intelligence

What is the issue?

Generative Artificial Intelligence systems are set to disrupt aspects of professional creative work, and have in fact already begun doing so. AI systems are now used for audiobook narration as a cost-cutting measure by large providers like Apple, though the market has so far indicated no preference for artificial narration over that provided by professional voice actors. Recent industry reports have highlighted instances where AI has also been used in the creation of cover art, a practice that stirs controversy with both creators and readers.

But the central concern at these early stages of the AI rollout is the question of how Large Language Models and AI are trained. Massive amounts of published writing are required to train AI. That writing is often protected by copyright, and the training of AI involves copying.

In September 2023, the U.S. publication The Atlantic published a searchable database containing the titles of books used to train a number of U.S.-based artificial intelligence projects. Many of those titles are by Canadian authors. 

TWUC Positions

NB: TWUC continues to study the growing AI market; therefore, our positions are evolving.

Authors reserve the right to refuse the use of their work for AI training.
AI training, including text and data mining, on published material without author permission is copyright infringement.
Governments must not provide AI companies new copyright exceptions for AI training and/or text and data mining.
AI training and the use of human-created content in AI outputs may only exist in a professionally licensed context.
All publishing contracts must include a clause defining AI training and text and data mining as author rights.
Authors must receive a larger share of any AI licensing payment negotiated by a publisher. Such licensing agreement must always be opt-in only.
All copyright pages in published books must include a warning that AI training is not permitted without specific licensing.
AI outputs must be clearly labeled as such. Presenting AI-generated work as human-created work is fraud.
Only human-created work may benefit from copyright protection.
Use of unlicensed word-for-word human-generated work in an AI output is plagiarism and an infringement of copyright.

What TWUC is doing

The Writers’ Union of Canada has been studying and discussing the question of artificial intelligence as related to creative professions for many years. Concerns that scanned books would be used for unlicensed “machine learning” were included in submissions made during the Google Books court action over a decade ago.

With the rise of Large Language Models and Generative AI, the issue of permission for use of creative work in the training of artificial intelligence is now very much in the media’s eye. The Union has been fielding many media and industry queries on the subject, are tracking two high-profile class actions in the United States aimed at securing permission and compensation for authors, and are discussing strategic next steps with our international colleagues.

Claims that AI training has accessed online “shadow libraries” — a.k.a. book piracy sites — to gather their huge datasets are particularly worrying. Regulators must not allow a new technology to legitimize illegal activity in this way.

We are aware there is a spectrum of opinion about the usefulness of AI in creative work. Our concerns, and those of our global colleagues are summed up well by the U.S. Authors Guild in an open letter they have addressed to AI developers:

Permission for the use of copyrighted material in generative AI programs must be sought and obtained.
Fair compensation must be paid for both past and ongoing use of creative works in generative AI programs.
Fair compensation must be paid for the use of creative works in AI output.

TWUC is grateful to our international partners, including the U.S. Authors Guild and UK Society of Authors, for their continued leadership on this issue. In September 2023, the Authors Guild launched a class action lawsuit (along with several high-profile U.S. authors) against the San Francisco-headquartered company OpenAI. As most AI development companies operate out of the United States, the initial use of the courts to rein in this damaging and illegal practice must take place there. 

Recent advocacy efforts include:

January 2024
As part of the ongoing public consultation on generative artificial intelligence and copyright, TWUC and fellow book industry associations submitted briefs reminding the Canadian government of the crucial importance of regulating the responsible development of artificial intelligence and ensuring effective copyright protection. Read the press release here.

The Union revised our Model Trade Book Contract publication with a clause aimed at reserving the author’s right to deny permission for AI training and the use of AI in publication. The publication is available as part of our Contracts Self-Help Package here, or accessible to members via the Portal. See the clause regarding AI below.

May 2024
TWUC was among a coalition of ten writing, publishing, and creator organizations from across the country that convened in Ottawa with parliamentarians and government officials on the opportunities and challenges generative artificial intelligence poses for Canada’s book industry. The organizations, spanning both French and English Canada, represent more than 300 book publishers and over 10,000 individual creators from across the country. The focal point of the reception was a book display featuring a selection of the thousands of Canadian titles that have been used without consent to train generative AI systems. The technology companies that have made use of these books did not seek the permission of the authors and publishers concerned and have made no effort to compensate rightsholders for this use. Read the press release here.

October 2024
Alongside a coalition of international creative industry organizations, TWUC mobilized a call to gather signatures on a statement on AI training. Thousands of writers, musicians, actors, artists, and other creators agree: Training generative AI on people's work without permission must not be allowed. Click here to learn more and sign the statement. 

November 2024
TWUC is studying a proposed agreement between select nonfiction authors, HarperCollins, and Microsoft to license full texts for the purpose of training artificial intelligence. As some TWUC members are affected, the Union has been directly consulted on the proposition. Read the press release here.

AI Licensing: What authors should know

Authors license their work for a number of reasons other than initial publication. Most contracts spell out all sorts of subsidiary rights licensing scenarios. The licensing of published work for AI training is now another of these subsidiary rights under the control of the author.

Licensing is not a barrier to innovation. Innovation does not mean free riding on the work of others, nor should it. Licensing preserves the integrity of copyright by giving creators an element of control over when and how their works are used. Licensing also provides legal certainty to good faith users, and keeps disagreements at the bargaining table where they belong, and not in court.

Will all authors need to sign up for AI licensing? Is it mandatory?
No. It is the individual choice of each writer to determine if they want to license their work. Even if a collective license solution is developed, TWUC will push to make sure in stays an opt-in design.
How does AI licensing interact with the recognized harms of generative AI to the writing profession?
Major generative AI tools were already trained illegally on creators’ work without a license to do so. This use is the subject of a number of lawsuits brought to stop the practice and at the very least convert training to licensed use. By converting unauthorized, unlicensed training to licensed uses, writers and publishers can set necessary restrictions on use and obtain fair payment. Writers and publishers can also decide not to engage — to not permit their works to be used at all.

In our consultations and submissions to Parliament, we have insisted that Government proceed with the most pressing and overdue revisions to the Copyright Act to make immediate repair to the cultural marketplace. TWUC recommends that the creative content uses required by AI must be negotiated within a market for licensing rather than through a damaging new exception to copyright.

The licensing solutions we are exploring give writers and publishers control over whether or not they want to permit certain output uses, and if they do, to earn adequate revenue from them.
What if I don't want to participate in a license? Does this mean that AI companies can simply use my work?
No. Not participating in a license does not give AI companies the right to use the works. That is an infringement of copyright.
If I am pursuing licensing, can I reach out to The Writers’ Union of Canada for assistance with AI licensing?
The Union is pleased to provide contract assistance to members who are reviewing contract offers from their publishers. It is essential that you understand the terms of the contract you are signing. We remind you not to let any publisher begin editorial revisions or production work until you have signed a mutually satisfactory contract. Contacts are negotiable — the clauses can, and generally should, be changed. For further details on contract advice, log in to the Portal.

What you can do

TWUC encourages all members to add their signature to the following statement on AI training:

“The unlicensed use of creative works for training generative AI is a major, unjust threat to the livelihoods of the people behind those works, and must not be permitted.”

Sign the Statement

Please Note: This statement is hosted by the non-profit group Fairly Trained, who are dedicated to seeing AI developers license their use of content for training. Fairly Trained is a supporter, as is TWUC, of the Human Artistry Campaign, which seeks to foreground the rights and professional needs of human creators as AI is widely adopted. This is part of the creative sector’s multi-pronged approach to the threat of AI, which also includes supporting legal action in the U.S. to challenge the unpermitted use of published work in AI training, and consulting closely with governments on the design and adoption of AI regulations. The statement has already received considerable press attention. If you at all uncomfortable with this approach, you need not sign the statement.

We encourage writers who are negotiating new contracts to include a clause regarding AI in their contract.  To ensure that your work is protected within the publishing context, we have recently added the following clause to the Union’s Model Trade Book Contract: 

Artificial Intelligence Clause
For avoidance of doubt, the Publisher has no rights to reproduce and/or otherwise use the Work in any manner for purposes of training or employing artificial intelligence technologies to generate text, or aural or visual elements, including without limitation, technologies that are capable of generating works in any form (“Generative AI Technologies”), in the same style or genre as the Work, in whole or in part, in any format (“AI Training Rights”), unless the Publisher obtains the Author’s specific and express prior-written permission to do so. Moreover, the Publisher does not have the right to sublicense AI Training Rights to others without the Author’s specific and express prior-written permission. The Publisher shall not include any text or other elements or enhancements in any edition of the Work that are created by or with the aid of Generative AI Technologies, or permit others to do so, without the Author’s specific and express prior-written permission.  

See links below for what you might do should you discover your work in one of these databases. Support the Union in our international advocacy and calls for effective regulation on this issue.

Why now?

Lots of generative AI companies train on creators’ work without a license to do so.
This is a major issue for many of the authors, artists, musicians, actors, and other creators whose work is being exploited by AI companies.
This is a critical time for creators around the world: in the U.S., there are multiple ongoing lawsuits that have been brought due to unlicensed training; in the UK, the government has said it would like to change copyright law and allow AI companies to train on copyrighted work without a license. Meanwhile, AI firms wishing to avoid meaningful regulation are relentlessly lobbying the Canadian government. Cultural sector organizations, like TWUC, are speaking out loudly to make sure artists’ rights are protected and preserved.