SFWA Sues Anthropic

The association for people who don't write science fiction is subpoenaed

SFWA has been subpoenaed for contact information from past and present members as part of its failing attack on textAI.

This summer, SFWA and other members of the Authors Coalition of America (ACA) have been involved in legal proceedings against Anthropic AI, which was trained on books downloaded from pirate websites.

The good news for authors is that Anthropic is now going to trial so the court can determine the amount of damages they will be liable for because of their use of our pirated books.

The court has decided this is not fair use, and a class-action lawsuit is now underway.

SFWA is now being subpoenaed for contact information from past and present members. This data will be used by lawyers to contact potential participants who fit the definition for this class-action lawsuit, so they can receive monetary damages.

You can find a partial list of pirated books used for Anthropic AI here.

Please note that not every book is eligible for this class-action lawsuit through the Authors Coalition. Per the court ruling on a definition for this case, an eligible text will have been registered with the United States Copyright Office within five years of publication and before being downloaded by Anthropic from LibGen or PiLiMi, or within three months of publication.

As this unprecedented case develops, SFWA will share more resources with members to support authors in decision-making around the defense of their work.

We strongly encourage all writers to share word of this lawsuit with writers outside of SFWA and other writers’ advocacy groups. You do not need to be an active member of any writing organization to participate in this legal action. If an author’s work falls under the aforementioned definition, they may be eligible to join this class-action lawsuit.

Thank you for your attention to this matter—and watch your inboxes for more.

Sincerely,

SFWA’s ACA Commission & Board of Directors

I won’t bore you with the details, but the mere involvement of SFWA in this lawsuit against Anthropic all but guarantees that the anti-Anthropic party is going to lose. In my experience, I have never come across a more inept and legally clueless organization than SFWA.

Fandom Pulse has the full analysis of the current state of the court case, which was already won in part by Anthropic. As it’s described, it looks like a complete non-starter:

The lawsuit's success depends on proving that AI training constitutes direct copying rather than transformative use. If successful, it could establish precedent requiring AI companies to license training data, potentially costing billions in damages and licensing fees.

Notice that it has already been established by the courts that AI generation is legally permissible transformative use. That was the much more important aspect. And I think there is absolutely no chance that it is going to be determined that training an AI on a book - which is exactly the same as reading a book - is going to be found any more harmful or financially damaging than an individual reading a pirated book.

Anthropic didn’t require more than one copy of a text in order to train an AI on it. Therefore, the greatest amount of damage to an author that one could possibly assign would be about $1 per book trained. Which, in my case, according to the link provided, would be around $32.

Which is $32 I would gladly contribute to the Anthropic legal fund if they didn’t already have a $65 million warchest with which to fight this ridiculous attempt to further lobotomize artificial intelligence.

Despite being a Life Member of SFWA since 1997, I stand entirely with Anthropic on this one.