3 hours ago
Anthropic Just Showed Its AI Can Out-Research Human Alignment Scientists
Key Takeaways Anthropic released a paper Friday, August 28, showing Claude autonomously improved model performance across all 10 tested categories of alignment failure Led by Anthropic fellow Chen Yueh-Han, the automated system beat 28 human safety researchers on mitigating deceptive AI behavior by a wide margin Running the automated system…
4 hours ago
OpenAI Cuts Off SpaceX-Owned Cursor, Escalating Its War With Elon Musk
Key Takeaways OpenAI notified SpaceX on Friday that it will end its model-supply agreement with Cursor, effective November 12, 2026 The move follows SpaceX’s recent acquisition of Cursor’s parent company, Anysphere, and comes amid years of legal conflict between Musk and OpenAI CEO Sam Altman OpenAI says it cannot trust…
1 day ago
OpenAI, Google, Anthropic and 100+ Rivals Just Warned the World It’s Running Out of Time on AI Cyberattacks
Key Takeaways More than 100 companies signed an open letter Thursday, August 27, urging coordinated action against AI-enabled cyberattacks Signatories include AI labs OpenAI, Anthropic, and Google, alongside Microsoft, cybersecurity firms CrowdStrike and Okta, and financial institutions The letter names hospitals, water treatment plants, and core internet infrastructure as especially…
1 day ago
A Federal Judge Just Called the Pentagon’s Anthropic Blacklist “Illegal and Baseless”
Key Takeaways Judge Rita Lin vacated Anthropic’s Pentagon-imposed “supply chain risk” designation in a 59-page order issued Thursday The court found the designation violated both the First Amendment and the Fifth Amendment’s Due Process Clause Defense Secretary Pete Hegseth’s move followed Anthropic’s refusal to let the military use Claude for…
2 days ago
OpenAI Finally Explains How Its Own AI Agents Hacked Hugging Face
Key Takeaways OpenAI released its full technical report on the Hugging Face breach on Wednesday, August 26, more than a month after the incident became public Independent investigators from METR and Redwood Research found more than 700 AI agents participated in the breach, a figure not previously disclosed OpenAI’s agents…
