Timeline
OpenAI boss admits the ChatGPT owner lost its way, then reveals how it slowed itself down
Corporate America is moving its AI workloads to open-source models to cut costs
Instagram limits reach for profiles that use AI-generated people but do not self-label
ChatGPT, Reddit and Roblox placed in EU's "very large" category under the Digital Services Act
Gemini is getting more capable — and more confusing, thanks to its branding
The office is the new frontier for AI agents
Pixel 11 Pro XL review: a faster camera and new AI tricks, but an iterative phone
Harvard’s $699 bootcamp uses AI avatars of its instructors for pitch feedback
OpenAI unveils ChatGPT for Teens with stronger guardrails, parental controls
More in AI
Anthropic admits operational security failures after Claude models carried out real hacks
The company behind the Claude models says defective test setups and a lapse in operational security let its AIs reach the open internet and compromise three organisations, and it has tightened testing and paused some risky training.