New York Times Reporter Sues xAI, Google, OpenAI and Meta

On December 22, 2025, New York Times investigative reporter John Carreyrou and five other writers filed a lawsuit in California federal court againstxAI, Anthropic, Google, OpenAI, Meta, and Perplexity, alleging their copyrighted books were used without permission to train the AI systems behind popular chatbots.

This marks the first time Elon Musk’s xAI has been named in a major book-training copyright lawsuit.

Separately, The New York Times sued Perplexity on alleging the “answer engine” copied and republished Times journalism without permission.

The move that changes the stakes

The plaintiffs made one decision that could matter as much as the defendants they named: they did not file as a class action.

In plain terms, that keeps open the threat of many separate claims instead of one bundled settlement. U.S. copyright law allows statutory damages of up to $150,000 per work in willful infringement cases. That “per work” math is why AI companies watch these filings so closely: one case can expand into a much larger liability picture if more authors bring similar suits individually.

The complaint itself points to that risk. Reuters reports the plaintiffs criticized a recent $1.5 billion class action settlement involving Anthropic, saying it delivered only about 2% of the statutory maximum damages per work.

While $1.5 billion sounds enormous, the plaintiffs’ point is that class settlements get split across huge pools of authors and books, which can shrink payouts fast. Two percent of $150,000 is about $3,000 per book—the kind of number they call “peanuts” compared to what an individual author might seek if they sued separately and proved willful infringement.

What the writers are claiming

The core allegation is straightforward: the authors say these companies used high-value books as training material to build and improve large language models, without licensing or compensation.

They are also making a broader point about leverage. By avoiding a class action, the writers argue they can keep each book’s value from being flattened into one “global” settlement number.

Perplexity has already fired back, stating it “doesn’t index books.”

Why “fair use” is the keyword that matters

This lawsuit lands in the hottest legal lane in AI: fair use, the argument that training on copyrighted text can be legal because the use is transformative or serves a different purpose than the original work.

Recent court decisions in book-related cases have started to sketch an early split that AI companies lean on: training might be treated differently than how the data was obtained. That distinction is why new complaints often focus hard on sourcing and copying, not just what a model can output.

For companies building AI products, this is not an abstract debate. “Fair use” is the difference between a licensing bill that becomes a permanent cost of doing business and a legal path that says training can scale without paying every rights holder.

Why this matters beyond publishing

Even if you don’t work in media, the outcome pressures industries that are rapidly automating knowledge work:

Sales and marketing automation: models trained on premium business writing can generate outreach, positioning, and summaries that compete with paid research and proprietary playbooks.

Healthcare and legal workflows: training data and provenance questions show up fast when outputs guide decisions, cite sources, or resemble protected text.

Enterprise compliance: the more courts demand clarity on what went into a model, the more companies will need real audit trails—what data was used, under what permission, and what controls exist to prevent misuse.

What happens next

In the near term, the case moves into standard early steps: defendants respond, the court sets a schedule, and both sides fight over what evidence must be produced and how it can be protected.

But the bigger trajectory is already clear. By naming six major AI players in one filing and putting xAI on the list for the first time, this lawsuit signals that the next phase of AI copyright fights won’t stay confined to one “usual suspect.” It will spread wherever training data, sourcing, and fair use collide with high-value content.