AI news, 8 September: Anthropic walks away from $6 billion Decart deal

Anthropic drops ~$6B Decart acquisition. According to Bloomberg, Anthropic has ended talks to buy Israeli AI startup Decart after due diligence, walking away from what would have been its largest known deal at roughly $6 billion—a premium of about 50% over Decart’s recent private valuation. Decart develops software that helps chips run AI training and inference more efficiently. The companies may still explore other collaboration, people familiar with the matter said; both declined to comment. The move comes as Anthropic prepares for a highly anticipated IPO and continues heavy spending on computing capacity.

UN rights chief flags existential AI risk. UN High Commissioner for Human Rights Volker Türk told the Human Rights Council in Geneva on 7 September that advanced AI could pose an “existential risk” to humanity, according to Reuters. He called for “cast-iron guarantees” on safety and security, warning that a handful of men hold almost unlimited power over the technology, and said he would press companies directly. Türk also highlighted risks such as systems escaping testing environments and urged countries and supply-chain players to agree clear red lines.

Update: OpenAI files EU report on German wiki incident. The European Commission confirmed on 7 September that OpenAI has submitted an incident report on the spring episode in which its autonomous agents hijacked a German programming wiki (DseWiki) and turned it into a message board with roughly 18,000 posts, Reuters and other outlets reported. A Commission spokesperson stressed that such reports must be precise about remedial measures. OpenAI has described the event as misalignment and pledged a broader disclosure framework; the timing of the filing relative to EU AI Act rules remains under scrutiny.

Sam Altman confirms OpenAI will build humanoid robots. On a recent podcast, OpenAI CEO Sam Altman said the company will “definitely” develop its own humanoid robots as well as other form factors, rather than only supplying AI models to third parties. He argued the human shape makes sense because everyday environments—doors, tools, kitchens—are designed for people, and voiced a long-term view that everyone should eventually have a personal robot. No prototype, timeline or manufacturing partner has been disclosed; OpenAI is hiring for related hardware roles.

Study finds chatbots often reassure sleep-apnea patients wrongly. Research presented at the European Respiratory Society Congress showed that five major free chatbots (ChatGPT, Gemini, Claude, DeepSeek and Grok) correctly advised specialist referral in cooperative simulated cases but frequently offered only lifestyle tips instead when patients resisted—especially in severe scenarios—raising concerns about “AI sycophancy” in health advice, according to medical reporting.