AI news · Saturday, August 29, 2026
OpenAI cuts off Cursor access citing Elon Musk’s history of violations
OpenAI announced it will pull support for Cursor, the popular AI-powered coding editor, on November 12, citing concerns that its new owner, SpaceX, cannot be trusted to follow terms of service. This follows a pattern of contract disputes involving Musk’s other ventures, including Twitter and xAI, which OpenAI alleges has previously scraped its models to train its own competing systems. While Cursor isn't accused of wrongdoing, it fell under a change-of-control clause when SpaceX acquired it, giving OpenAI a legal window to exit. The breakup leaves developers who rely on OpenAI’s top-tier models for coding in a lurch, even though OpenAI says it plans to support them through the transition.
Legal trouble is also brewing for Anthropic, which is currently being sued by Sony Music and Warner Chappell for alleged mass-scale intellectual property theft. The music publishers claim Anthropic illegally downloaded and scraped millions of copyrighted tracks to train its Claude models, even pulling content from known piracy hubs like Library Genesis. The plaintiffs are seeking up to $150,000 in damages per composition, a figure that could climb into the billions if the case gains traction. This is a heavy blow for Anthropic, which already settled for $1.5 billion in a similar previous case involving copyright infringement through pirated content.
Inside the office, companies are trying to figure out how to stop the 'AI slop' epidemic. Andrew Wang, the CEO of mortgage-tech startup Valon, has banned new hires from using AI for basic tasks, arguing that relying on models too early prevents workers from actually learning their jobs. The policy is paying off financially—annualized token spending is projected to drop from roughly $20 million to $5 million—but the main goal is ensuring employees can spot when an AI is hallucinating before they pass off bad work to others. This sentiment is gaining ground as companies realize that 'meat proxies'—humans who share unchecked AI outputs without oversight—are now a significant drag on productivity. Meanwhile, over at Nvidia, the company is betting its long-term future on system-level data orchestration rather than just raw GPU power. As competitors like Google and Amazon build their own chips, Nvidia is doubling down on its 'Vera Rubin' architecture, which aims to optimize the traffic flow between storage and memory, proving that in a post-GPU-monopoly world, moving data efficiently is the new competitive moat.
The quick hits
- OpenAI is cutting off coding startup Cursor on November 12, citing deep distrust of owner Elon Musk’s past contract violations at Twitter and xAI.
- Sony Music and Warner Chappell sued Anthropic for allegedly using pirated music to train Claude, seeking up to $150,000 per song in damages.
- Valon’s CEO banned new hires from using AI to perform basic work tasks to prevent 'meat proxy' behavior and ensure employees actually understand their roles.
- Nvidia is shifting focus from raw GPU power to complex data orchestration, hoping its Vera Rubin architecture will keep it ahead of Amazon and Google.