AI This Week: Lawsuits, a Cutoff, and a Breach

The week of August 24 to 30 was defined by the business and legal side of AI rather than new models. The three major music publishers sued Anthropic in a case with billion-dollar exposure, OpenAI moved to cut Cursor off from its models after a competitor acquired the tool, malware siphoned active Claude sessions from users, and enterprise buyers kept shifting toward cheaper models over frontier flagships. Here are the stories that mattered and why they matter for anyone building on or investing in AI.

Sony and Warner Sue Anthropic Over Claude

It was announced that on August 28, Sony Music Publishing, Warner Chappell Music, and dozens of affiliated publishers, 35 entities in all, sued Anthropic in the Northern District of California, naming CEO Dario Amodei and co-founder Benjamin Mann personally. The 48-page complaint alleges Anthropic built Claude using tens of thousands of copyrighted song lyrics obtained through piracy and scraping.

The specifics are pointed. The publishers claim Anthropic torrented lyrics and sheet music from pirate archives like Library Genesis, scraped licensed lyric sites such as MusixMatch and LyricFind against their terms, and “cleaned” the text by stripping out copyright notices, which the complaint frames as deliberate concealment. They allege Claude can reproduce near-identical copies of protected lyrics, seek statutory damages up to $150,000 per composition, a potential multi-billion-dollar figure, and want a jury trial. Anthropic said it disagrees with the claims and will defend itself in court. The case follows Anthropic’s $1.5 billion settlement with authors in 2025 over the same pirate archives, and with all three major publishers now aligned, it marks a decisive phase in AI’s copyright reckoning.

OpenAI Moves to Cut Off Cursor After SpaceX Deal

In a striking sign of how competitive the AI coding market has become, OpenAI notified the maker of Cursor on August 28 that it will wind down the contract supplying its models to the popular coding tool, with a proposed shutdown on November 12, 2026. The trigger is a change-of-control clause: SpaceX acquired Cursor’s parent company Anysphere in a reported $60 billion deal, and OpenAI invoked the clause rather than keep supplying a now-rival-owned product.

Cursor CEO Michael Truell confirmed the notice, saying OpenAI models account for only about 5% of Cursor’s traffic and that the company is in talks to resolve it. Anthropic, meanwhile, is positioned to step in as the convenient alternative, a reminder that model access is now a strategic lever, not just a supply agreement.

Malware Hits Claude Users

Security was the week’s third thread. Anthropic began signing out affected Claude users, wiping saved payment methods, and refunding unauthorized charges after infostealer malware on users’ own PCs siphoned active Claude login sessions and consumed usage limits without permission. The company named several malware families, including Vidar, LummaC2, StealC, and RedLine, as the culprits.

The important nuance: this was not a breach of Anthropic’s systems but malware on individual users’ machines stealing session tokens, a growing attack pattern as AI accounts become valuable targets. Still, it pushed the company into an unusual round of forced sign-outs and refunds.

Cheaper Models Keep Winning

The quieter but structurally important story is economic. New reporting showed corporate spending on Anthropic’s flagship Fable 5 has plateaued at roughly 11% of spending on the company’s tools, as businesses conclude cheaper models handle most workloads. Anthropic’s lower-priced Opus 5 has already passed Fable 5 in business spending, and GPT-5.6’s lower price helped OpenAI regain momentum.

The takeaway for anyone budgeting AI: the reflex of defaulting to the most capable frontier model is fading, with analysts suggesting flagships may increasingly serve as showcases of progress rather than the everyday enterprise default. Matching each task to the cheapest model that clears the bar is now the smart play.

Also Worth Knowing

• Google ships Gemini Omni 1.1 Flash: the Aug 27 update to Google’s Omni video model adds 40-second clip extensions, first- and last-frame control for consistency, and cheap 360p drafts at $0.03 per second that upscale to 4K. Google says it tops the Text-to-Video Arena at 1,515 points, a vendor-reported figure rather than a confirmed ranking.

• Containment transparency questioned: an independent assessment by Guidelight AI standards found major labs published little on how they would contain a model that subverts human control, with OpenAI scoring highest for having paused workloads after past incidents.

• Claude Cowork gets a browser: Anthropic added a built-in web browser to Claude Cowork, letting it search the public web directly without a separate extension.

• Regulation tightens: California and New York now require AI safety frameworks, and a proposed federal bill calls for technical mechanisms to shut down rogue models.

The Week in One Line

AI’s center of gravity shifted from the lab to the courtroom and the balance sheet: a landmark copyright fight, a supply cutoff driven by a rival’s acquisition, a user-security scare, and a steady march toward cheaper models. The models keep improving, but this week the decisive forces were law, money, and trust.