Authors sue OpenAI, Meta, Google and others over AI training on pirated books
A new lawsuit from a group of writers claims Anthropic, Google, OpenAI, Meta, xAI, and Perplexity used pirated copies of books to train AI systems without permission. Among the plaintiffs is Theranos whistleblower and Bad Blood author John Carreyrou.
This fight didn't start here. An earlier case against Anthropic set the stage: a judge indicated that training on copyrighted books may be lawful, but obtaining those books through piracy is not. That case ended in a large payout-and fresh backlash from authors.
What happened in the earlier case
Anthropic agreed to a $1.5 billion settlement. Eligible writers could receive around $3,000 each.
Many authors called it unfair. They argue the payment doesn't match the value of their work or the scale of the alleged infringement.
What this new lawsuit says
The new filing claims the prior deal benefits AI companies more than creators. According to the complaint, the proposed Anthropic settlement "seems to serve (the AI companies), not creators."
It adds: "LLM companies should not be able to so easily extinguish thousands upon thousands of high-value claims at bargain-basement rates, eliding what should be the true cost of their massive willful infringement."
Why this matters for working writers
Your backlist, your advances, and your licensing leverage are on the line. If courts push for cleaner data sources and real licensing, the upside for authors could be meaningful. If not, expect more low payouts and vague promises.
The outcome could set norms for how books and other copyrighted materials are used in AI training-what's permitted, what must be licensed, and how compensation flows back to creators.
What could change next
- Clearer rules: Courts may draw sharper lines between lawful training and unlawful sourcing.
- Licensing models: Deals with publishers, agencies, and collectives could expand-rates and terms will matter.
- Discovery pressure: AI firms may face deeper scrutiny over data provenance and content filters.
- Compensation structure: Flat per-author checks may give way to tiered payouts tied to usage and value.
Action steps for authors right now
- Register your works. It strengthens your position in any infringement claim. If you need the official link, see the U.S. Copyright Office's registration portal: copyright.gov/registration.
- Keep clean records: publication dates, ISBNs, editions, excerpts, and licensing agreements. This speeds up claims and audits.
- Track eligibility windows for settlements and opt-outs. Don't miss deadlines-put reminders on your calendar.
- Update contracts and website terms to address AI training, dataset use, and text/data mining rights.
- Join or coordinate with author groups that monitor infringement and negotiate on behalf of writers.
- Use unique text identifiers or controlled excerpts on public sites to help detect misuse without giving away full manuscripts.
Bottom line
Authors are pushing for two things: lawful sourcing and fair pay. This case aims to test both.
If courts demand verifiable data provenance and meaningful licensing, writers will have a stronger seat at the table. If not, expect more low, one-size-fits-all checks that don't reflect the value of your catalog.
Related resource for working writers
If you're experimenting with AI in your workflow, here's a vetted list of tools used by copywriters: AI tools for copywriting. Use them wisely-and keep your rights protected.
Your membership also unlocks:
