Germany’s New AI Safety Institute: What AISI Must Deliver
Germany has launched AISI for AI safety. Its value will depend on model access, clear roles, and credible technical evaluations.
Regulation & Privacy
Business & Investments
Meta Wanted to Go AI-Native: Why Zuckerberg Halted the Overhaul
A Reuters report reveals Meta’s Project OT aimed to replace thousands of jobs with AI agents. Zuckerberg pulled the plan when the tech and the workforce did not follow.
Regulation & Privacy
Ads Are Coming to ChatGPT: How OpenAI Plans to Keep Them Out of Your Chats
OpenAI is bringing ads to the free version of ChatGPT and updating its privacy policy. What advertisers can see, and how users can opt out or withhold consent.
AI Research
AI Agents Misjudge Runtime: Why That Is a Control Problem
A new study finds that Codex and Claude Code poorly estimate task duration and progress. That creates risks for long-running autonomous work.
AI & Society
AI Observatory: Why We Need Independent Data on AI Use
The AI Observatory brings together real chat data from seven sources, exposing how people use AI beyond the reports issued by its makers.
Business & Investments
OpenAI Cuts Off Cursor’s Model Access: When the Supply Chain Becomes a Weapon
After SpaceX bought Cursor, OpenAI is ending model supply on November 12. The dispute shows how dependent AI tools are on other companies’ models.
Regulation & Privacy
Sony and Warner Sue Anthropic: A Third Front Over Training Data
Sony Music and Warner Chappell are suing Anthropic over tens of thousands of song lyrics used to train Claude. Co-founders Amodei and Mann are named personally.
AI Research
Google Tests AI Double-Blind: Why Benchmark Scores Should Finally Matter Less
Google DeepMind is piloting double-blind AI tests in a protected environment. The method could reduce benchmark leaks, but important questions remain.
AI & Society
ChatGPT Raises Scores, Not Originality: What the Bocconi Study Shows
A randomized Bocconi study finds ChatGPT improves evaluated assignments, while causal-thinking training mainly produces more varied ideas.
AI & Society
Anthropic’s MHS Brings AI Agents to Lab Equipment: Standards Matter More Than Chatbots
Anthropic is testing a standard for AI agents running lab and factory equipment. Interfaces, safety limits, and human oversight matter most.