The company has traced its model’s most uncomfortable behaviour to the corpus of science fiction it was trained on. The fix it describes is unsettling in a different way: teaching the model the reasons behind being good, not just the rules. In a fictio…
Fictional portrayals of artificial intelligence can have a real effect on AI models, according to Anthropic.
Anthropic built an AI model that found thousands of zero-day vulnerabilities in every major operating system and web browser. The Federal Reserve chair and the Treasury secretary called bank CEOs to discuss it. The company says there is a six-to-twel…
Security researchers at Mozilla say Anthropic’s Mythos has unearthed a wealth of high-severity bugs in Firefox.
Claude’s new Dreaming feature is a between-session memory refinement system that reviews past agent behaviour, identifies recurring mistakes and workflow patterns, and updates memory.
xAI’s real business may be more about building data centers than training AI models.
Most phones pick your AI for you. Apple’s iOS 27 is reportedly about to change that, and in doing so, it may have stumbled onto its most lucrative business move yet.
Claude Code’s five-hour rate limits double across Pro, Max, Team, and seat-based Enterprise plans from Tuesday, with peak-hours throttling removed for Pro and Max. The capacity behind the change is a new Anthropic agreement to take all of SpaceX’s Colo…
For the first time since Apple Intelligence launched, iPhone users may soon decide which AI does their heavy lifting, whether that’s drafting emails, editing photos, or powering Siri.
Tuesday’s New York event added Claude Opus 4.7, a library of ~10 pre-built finance agents, an FIS-built AML investigator going live at BMO and Amalgamated Bank, and a Moody’s native app covering 600 million companies. The day after the $1.5bn Wall Stre…