Anthropic Settles Major Copyright Class Action Filed by Authors

Anthropic co-founder and CEO Dario Amodei. Photo Credit: TechCrunch

Just months out from a scheduled trial, Anthropic has officially finalized the terms of a settlement in the class-action copyright lawsuit levied against it by authors.

Last time we checked in on the high-stakes case, the presiding judge rejected Anthropic’s stay request. And as noted, the development left the AI giant months away from a planned December trial involving many authors whose protected works were allegedly pirated as part of Claude’s training process.

That reality rather unsurprisingly prompted an appeal. Therein, Anthropic’s legal team emphasized “the unprecedented and erroneous certification of possibly the largest copyright class action ever,” pointing to as many as seven million claimants as well as “hundreds of billions of dollars in potential damages liability.”

Now, we needn’t continue tracking the push for appellate-court relief or developments concerning the formerly quick-approaching trial.

In a new filing, the litigants told the same court that they’d “executed a binding term sheet intended to memorialize the terms of a proposed class settlement.” Additionally, the parties informed the district court of the settlement.

Judge William Alsup subsequently set a September 5th deadline for the authors and the Google-backed AI company to submit a motion for preliminary approval. Along with a scheduled September 8th hearing, this motion should provide details about the agreement’s specifics.

Furthermore, Judge Alsup underscored that the December 1st trial start date will remain in effect “in the event that the settlement craters.”

In other words, there’s a clear-cut incentive for Anthropic to drive the deal into the endzone.

Shifting the focus to the music space, it remains to be seen whether the same incentive will prompt a settlement in major music publishers’ own ongoing (and seemingly expanding) copyright suit against Anthropic.

Notably, the latter complaint involves alleged infringement both in generative models’ training processes and in their outputs; the authors’ action centers on training and piracy.

(If “the outputs seen by users had been infringing,” Judge Alsup previously wrote of the authors’ class action, the plaintiffs “would have a different case.”)

Running with the detail, it was only in June that the court handed a key win to Anthropic in the authors’ case, finding that the actual process of training LLMs on protected works without authorization was (and presumably is) “quintessentially transformative.”

Furthermore, the judge also determined that Anthropic’s breaking down and scanning physical books to build a digital “research library” likewise constituted fair use – with subsequent copies made expressly for training purposes falling under the banner as well.

Stated differently, the victory for authors here stemmed from Anthropic’s alleged piracy – not a triumph on the ever-pressing fair use question.

It appears safe to say that the company won’t allegedly pirate books moving forward, and despite contrasting legal-system takes on fair use in gen AI, the point will certainly prove significant when it comes to the possibility of future cases.

Two important closing notes: It will, of course, be interesting to see how the settlement affects copyright litigation involving fellow AI behemoths. Plus, music publishers’ Anthropic action is proceeding at a comparatively relaxed pace.

Earlier in August, Judge Eumi Lee granted jointly sought extensions that, among other things, moved the planned trial date to September 28th, 2026. Notably, given prior settlement-talk rumblings, the litigants were seeking longer delays than those the court ultimately granted.