Version 3

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 distinguishing AI training (fair use) from direct piracy (likely not) [2][1]. Roughly 350 authors who opted out filed a separate lawsuit requesting a jury trial, seeking higher damages [4]. In a parallel case, a federal court rejected Google's key DMCA claims against web scraper SerpApi, holding that anti-scraping technology does not constitute copy protection under the statute; Google filed the suit in late 2025 calling SerpApi 'parasitic' and is now appealing [7][10][11]. OpenAI changed ChatGPT's behavior to decline direct mimicry of named authors' styles, and artist Elmer Saflor has sued an AI meme generator for commercially exploiting his viral comic [12][13].

The Anthropic settlement produced the first major judicial distinction between training-data fair use and direct piracy, but without creating binding appellate precedent; the opt-out authors pursuing a jury trial may force courts to establish binding standards on damages. The Google v. SerpApi ruling, if it survives appeal, would confirm that companies cannot use DMCA anti-circumvention claims to control access to publicly visible web data — a finding with significant consequences for AI companies and data aggregators that rely on scraping.

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]. The case produced a notable legal distinction in earlier proceedings: 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 [2]. The settlement resolved the class action without creating binding appellate precedent, but the fair use / piracy framing is one that other courts and AI developers will likely reference. Authors' reactions were mixed: some welcomed the court's finding that direct piracy is actionable, while others argued that individual payouts of approximately $3,000 per work were too low for the commercial value Anthropic extracted, and that attorney fees consumed too large a share of the fund [3][2].

Not all class members accepted the result. 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 and has continued questioning whether the settlement meaningfully changed the AI copyright landscape for authors [5][6].

A parallel dispute over web scraping produced a ruling with implications beyond AI training data. Google sued SerpApi in late 2025, calling its commercial scraping of Google search results 'parasitic' and arguing that its anti-scraping technology constitutes copy protection under the DMCA [7][8]. SerpApi responded with a motion to dismiss, arguing that Google does not own publicly accessible search results and that it is in the right [9]. A federal court rejected Google's key DMCA claims, ruling that anti-scraping technology does not function as copy protection under the statute [10]. Commentators characterized the outcome as the judge declining to let Google 'DMCA its way out of being scraped' [11]. Google announced it will appeal.

On the product side, OpenAI changed ChatGPT's behavior in late July 2026 to decline direct requests to write in the style of named authors — a restriction previously applied only to living writers but now extended to deceased ones like Hemingway and Dickens [12]. ChatGPT now offers to capture an author's 'broad qualities' while maintaining its own voice. Separately, artist Elmer Saflor filed a copyright infringement suit against Memes Apps LLC, alleging the company's paid AI meme-generation service reproduces his 'Running Away Balloon' comic — which went viral in 2017 — without authorization [13]. The case tests whether widespread unpaid viral sharing forecloses a creator's right to demand licensing when an AI platform commercializes the same work.