AI Best Practices for Authors: Navigating Ethical and Legal Boundaries

The Authors Guild has taken a proactive role in addressing the implications of artificial intelligence for writers. Recognizing the need for informed decision-making, the Guild first published best practices for AI use in February 2024. Their newly updated best practices, informed by extensive conversations with writers and deliberations by the Authors Guild Council, aim to provide enhanced context and clarity. These guidelines are intended to empower members of the writing community to make their own ethical and informed choices regarding the adoption or rejection of AI technologies in their work.

Navigating the use of AI in writing involves understanding several critical risks. Firstly, AI-generated text is not copyrightable due to the absence of original human authorship. Any AI-generated content included in a work submitted for copyright registration must be disclosed and disclaimed. Failure to do so can be considered fraud on the Copyright Office and may invalidate the registration. Secondly, book contracts typically include representations and warranties that the manuscript is original to the author. AI-generated material does not meet this definition of 'original authorship' and its inclusion can constitute a breach of contract, potentially allowing publishers to terminate agreements. Thirdly, there's a risk that LLMs, having ingested a book during training, may regurgitate its words, plot, or characters. AI outputs can inadvertently include recognizable elements from other authors' copyrighted works, despite claims to the contrary by AI companies. Finally, writers are ultimately responsible for the manuscript they deliver to a publisher or self-publish, regardless of AI assistance.

The inclusion of AI-generated text in a manuscript can have significant contractual repercussions. Most book contracts contain a clause where the author represents and warrants that the submitted manuscript is their original creation. Since AI-generated material is not considered original authorship, its inclusion can directly violate this warranty, constituting a breach of contract. In such cases, publishers may reserve the right to terminate the publishing agreement. It is therefore crucial for authors to be transparent with their publishers about any AI outputs incorporated into their work. If a publisher is registering the book on the author's behalf, they must be informed about the use of AI to ensure the application is completed correctly and honestly.

A significant concern when using public AI chatbots is the potential for input material to be used for training the models. Uploading drafts, outlines, research notes, or interview transcripts to these platforms can inadvertently expose confidential information. This practice may compromise an author's future rights and could conflict with exclusivity clauses in publishing contracts. While most consumer AI tools offer settings to opt out of training on user inputs, these settings can sometimes be difficult to locate or understand, and their long-term effectiveness is not always guaranteed. Authors must be vigilant about protecting their proprietary information when interacting with AI systems.

To navigate the use of AI responsibly, authors are advised to adhere to several best practices. Crucially, if claiming authorship, AI should not be allowed to replace an author's unique voice, thinking, and creativity; the words and thoughts must remain their own. When using AI for research or as a sounding board, it should be treated as background material only, with AI-generated output strictly avoided in the final work, akin to preventing plagiarism from any third-party source. All AI outputs must be thoroughly reviewed and fact-checked for accuracy and potential biases, including gender, racial, and socioeconomic prejudices. Any AI-generated text incorporated into a work must be disclaimed in copyright applications, and publishers must be informed if they are handling the registration. Authors must ensure AI use does not violate contractual warranties of originality. Transparency with agents and publishers about AI incorporation is essential, and discussions should be held to ensure alignment, especially if the AI use is integral to the work itself. Authors should also inquire about specific publisher guidelines on AI use and ensure any third-party collaborators on manuscripts adhere to strict AI usage and disclosure clauses. Public disclosure of substantial AI text generation is encouraged to maintain reader trust and uphold the value of human authorship.

The Authors Guild encourages authors to consider their solidarity with professional creators in related fields who are also grappling with the impact of AI, such as voice actors, narrators, translators, and illustrators. Authors should be mindful of whether their use of AI might displace human professionals. When AI is used instead of hiring a human, the Guild recommends seeking AI programs that were trained on licensed content rather than those built on copyright infringement. The Guild also advocates for authors to continue supporting human translators, who play a vital role in reaching wider audiences. Furthermore, authors should respect their fellow writers by refraining from using generative AI to deliberately copy or mimic unique styles, voices, or distinctive attributes of other writers' works in a way that devalues their contributions or attempts to profit unfairly from them. Such actions could lead to claims of unfair competition or copyright infringement.