Version 2

A federal judge approved the $1.5 billion Bartz v. Anthropic settlement in July 2026 — the largest copyright settlement in U.S. history — with the court holding that AI training on books constitutes fair use but direct piracy likely does not [1][3]. Roughly 350 authors who opted out filed a separate copyright suit in May 2026 requesting a jury trial, testing whether they can recover more than the ~$3,000-per-work class payout [4]. Google separately lost a DMCA suit against web scraper SerpApi but is appealing, arguing its anti-scraping technology functions as copy protection under the DMCA [6]. ChatGPT now refuses to directly mimic named authors' styles, and artist Elmer Saflor is suing an AI meme generator for commercially exploiting his viral comic without authorization [7][8].

The Anthropic settlement produced the first major judicial distinction between training-data fair use and direct piracy, but the opt-out authors pursuing a jury trial could force courts to produce binding precedent rather than a negotiated resolution. Google's DMCA theory — that anti-scraping infrastructure constitutes copy protection — if upheld on appeal, would extend copyright-adjacent protections to data-collection pipelines and reshape how AI companies can legally gather training material from the web.

A federal judge granted final approval in July 2026 to a $1.5 billion settlement between Anthropic and a class of authors in Bartz v. Anthropic, the largest copyright settlement in U.S. history [1][2]. The case's earlier proceedings produced a notable legal distinction: the court found that Anthropic's use of books to train its AI models constituted fair use, but that the company's direct piracy of copyrighted works likely did not [3]. The settlement closed the class action without creating binding appellate precedent, but the fair use / piracy framing is one other courts and AI developers will likely reference as litigation continues.

Not all class members accepted the result. Authors who opposed the settlement argued that individual payouts of approximately $3,000 per work were too low for the commercial value Anthropic extracted from their writing and that attorneys' fees consumed too large a share of the fund [3]. Roughly 350 authors formally opted out; a group among them filed a separate copyright lawsuit in May 2026 requesting a jury trial, signaling an intention to pursue higher damages than the class settlement provided [4]. Writer Beware, which tracked the settlement skeptically throughout, covered the opt-out litigation under the heading 'The Opt-Outs Strike Back' [5].

A separate legal dispute is testing whether anti-scraping technology constitutes copyright protection. Google lost a DMCA suit against SerpApi — a commercial service that sells access to an API built on scraped Google search results, including knowledge panels containing licensed third-party content — but announced it will appeal [6]. Google's theory is that its anti-scraping measures function as copy-protection mechanisms whose circumvention violates the DMCA; SerpApi's counter-argument is that Google does not own the internet and that publicly accessible search results are not DMCA-protected. Reddit is also participating in the legal effort against AI web scraping alongside Google, though it is not a direct party to the SerpApi dispute [6].

OpenAI separately changed ChatGPT's behavior in late July 2026 to decline requests to write directly in the style of named authors — a restriction previously applied to living writers like Stephen King and J.K. Rowling but now extended to deceased ones like Hemingway and Dickens [7]. ChatGPT now offers to capture an author's 'broad qualities' while maintaining its own voice. On the same day, artist Elmer Saflor filed a copyright infringement lawsuit against Memes Apps LLC, alleging the company's paid AI meme-generation service reproduces his 'Running Away Balloon' comic without authorization [8]. The lawsuit turns on whether widespread unpaid viral sharing of a work forecloses the creator's right to demand licensing when an AI platform commercializes that same work.