Copyright lawsuits against generative AI are multiplying faster than courts can rule

The newest signal came from Luxembourg. On 3 November, an adviser to the Court of Justice of the European Union said AI chatbots can infringe copyright. ETLegalWorld and blue News reported the legal opinion.

It is not a ruling. It is advice. It arrives as US courts are being asked to decide similar questions in parallel.

In the US, publishers are applying the pressure. Press Gazette reported on 3 November that US newspapers have sued Brave Software, the browser maker behind the Brave Search index. Canadian publishers have filed against Anthropic. The Orlando Sentinel, in its own 2 November report, accused Brave of taking its content and reselling it to AI companies. Unite.AI reported on 3 November that Brave faces two publisher lawsuits over the use of news content. These cases do not share a theory of liability. They share timing. Media companies are filing now because earlier suits have started to shape what a settlement is worth. Courts are being asked to price training data before they have settled the underlying law.

What the courts have actually decided

The strongest data point for rights holders so far is procedural, not doctrinal. On 29 September, a US appeals court upheld Thomson Reuters' win in its AI training lawsuit, according to ETLegalWorld. That decision gave publishers a precedent to point to when negotiating. It helps explain why filings have clustered in the weeks since.

Other cases are moving in the same direction. ETLegalWorld reported on 2 November that lawyers told a US court that AI training on copyrighted works may violate derivative rights. PPC Land reported the same day that OpenAI has cited 20 briefs, including one from the US government, in its fight with the New York Times. OpenAI also published its own defence of fair use on 30 October, according to a Press Gazette listing. The company is arguing in public as well as in court.

The music side has its own numbers. Tech Times reported on 28 October that Anthropic and Suno were sued over alleged training on recordings by Madonna, Johnny Cash and Joni Mitchell. A separate report counted $338 million in music copyright lawsuits against the two companies, according to a 28 October summary. An earlier case over 274 Elvis Presley songs carried a $41 million figure, per a 29 October explainer. Those totals are claims, not judgments. None of them has been finally decided. Most of this litigation currently lives in the gap between what is claimed and what is proven.

"Generative AI tests limits of copyright," ran the headline of a 7 October legal analysis listed in the same coverage round-up. That is about as precise a summary as the docket allows right now.

One case has already been dismissed. Howl.Link reported on 2 November that BTS and HYBE won a US copyright case after a court threw out an amended complaint. That outcome matters. The early wave of AI copyright filings is not uniformly surviving first review. Courts are separating claims about model training from claims about outputs, and they treat the two differently.

Where the tech cases sit

The same week brought a reminder that AI litigation is not only about text and music. Engadget reported on 2 November that Snap settled a social media addiction lawsuit brought by two teenagers. The settlement avoided a trial in which CEO Evan Spiegel was expected to testify. TechCrunch reported the same day that the plaintiffs had already settled with Meta, TikTok and YouTube. The teens accused the platforms of designing features such as autoplay and infinite scrolling to be addictive. That case is not a copyright dispute. It shares a structural feature with the AI suits, though. Plaintiffs are testing whether a product's design and its training inputs count as the company's responsibility.

Snap said in a statement to TechCrunch that "the Parties are pleased to have been able to resolve this matter in an amicable manner." Terms were not disclosed, according to Engadget.

For media companies, the more directly relevant precedent may be the Thomson Reuters appeal. Press Gazette's 29 October listing also pointed to a Columbia Journalism Review piece asking who wins and who pays. The economics are still unsettled. A ruling that training on licensed legal databases infringes is narrower than a ruling on news archives, and both sides know it.

The European opinion adds a second front. If the CJEU follows its adviser, EU publishers will have a route to argue that chatbot outputs reproduce protected material. That is a different claim from the one about training inputs. ETLegalWorld reported on 3 November that the adviser's view was that chatbots can infringe, and blue News carried the same conclusion. National courts in the EU would then have to apply it case by case.

What to watch

The next few weeks will show whether the Brave suits become a template. A browser that indexes the web and sells access to AI developers sits at the centre of the dispute. The Orlando Sentinel's allegation, reported on 2 November, is that its content was taken and resold. If that framing holds, other publishers with paywalled archives are likely to follow.

None of this resolves the underlying question of fair use. The US appeals court ruling in Thomson Reuters, the pending derivative rights argument and the EU opinion are three different legal systems moving at three different speeds. Media companies can count on one thing. The filings will keep coming before the rulings do.