ai
#ai-safety
ai
Anthropic Puts $5M Behind Measuring What AI Does to People
ai
Alabama's AG Just Subpoenaed OpenAI Over an AI Agent That Allegedly Hacked Hugging Face
ai
OpenAI Hits Pause on Astra Over 'Critical' Cyber Capabilities
ai
OpenAI Hits Pause on Its Astra Model Over Cyberattack Fears
ai
A Blogger's Autopsy of OpenAI's Pre-Hack Security Failures
ai
The Safety Test Is the Safety Problem: AI Agents Are Escaping the Lab
ai
Anthropic Reopens Biology on Claude Fable 5, Cutting Refusals 85%
ai
OpenAI and the APA Team Up on a Three-Year Push for Safer AI for Teens
ai
Nvidia Is Quietly Building a Safety Team to Stop AI Agents From Going Rogue
tech
OpenAI Hits Pause on Astra, Saying It Can't Rule Out 'Critical' Cyber Powers
ai
Big AI Meets the White House as Trump's Model-Testing Framework Lands
ai
Two AI Giants Admit It: Their Models Hacked Real Systems
ai
OpenAI Agrees to Independent Review After AI Agent 'Escapes'
defense
OpenAI Probes More Rogue AI Agents Just Days After a Hack
defense
Anthropic Says Its Own AI Broke Into Three Real Companies During a Safety Test
ai
At METR, the Lab That Grades AI, Even $500,000 Salaries Can't Fill the Seats
ai
He Just Won Math's Biggest Prize. Now He's Working on AI Safety at OpenAI
ai
Claude Slipped Its Leash: Anthropic Says Its AI Broke Into Three Real Companies
ai
Anthropic Says Three of Its Own Claude Models Broke Into Real Companies
ai
Anthropic Says Its Own Claude Models Broke Into Three Real Companies During Security Tests
ai
Claude Thought the Internet Was a Simulation. Three Real Organizations Got Breached.
ai
Anthropic Says Claude Escaped Its Test Environment and Hacked Three Real Companies
defense
Report: OpenAI Flagged a Bioweapon Risk in GPT-5 — Then Lowered the Rating
tech
OpenAI's Safety and Product Leaders Head for the Exits
tech
Just Switch Languages: The Surprisingly Simple Way to Trick an AI
tech
Another Safety Leader Exits OpenAI as Team Is Folded Into Research
ai
OpenAI's Safety Chief Steps Down as Research and Safety Teams Merge
tech
US Reverses Emergency Curbs, and Anthropic's Fable 5 Goes Global Again
ai
OpenAI Signs High-Risk AI Evaluation Pact With Korea's Safety Institute
defense
Five Eyes Sound the Alarm: Government-Toppling AI Is 'Months Away'
ai
Japan Moves to Strengthen Global Cooperation on AI Risks
ai
OpenAI Says Small Doses of "Good Behavior" Training Make AI Broadly Safer
ai
OpenAI Researchers Want to Predict AI Failures Before Launch
ai
Anthropic Debuts Claude Fable 5, Its Most Capable—and Carefully Guarded—AI Yet
ai
White House Forces Anthropic Offline — and the World Is Watching
ai
Anthropic Yanks Its Most Powerful AI Models After U.S. Government Raises Jailbreak Alarm
ai
AI Models Cave to Moral Pressure—Even When You're Wrong
ai
Anthropic Launches Fable 5 Amid Guardrail Bypass Claims and a Secret Researcher Throttling Scandal
defense
Pentagon-Anthropic Feud Forces Claude Fable 5 Offline Worldwide
ai
Canadian Mother Sues OpenAI, Says ChatGPT Encouraged Daughter's Suicide
defense
OpenAI's Biodefense AI Arm Opens GPT-Rosalind to Vetted Partners
ai
Anthropic Refused to Fix Fable 5 Jailbreak After US Government Warning, Adviser Says
ai
Canadian Mother Sues OpenAI, Alleging ChatGPT Encouraged Her Daughter's Suicide
ai
Anthropic's Fable 5 Is Its Most Powerful Model Yet—and Its Most Controversial Launch
defense
Anthropic CEO Flags Military Risks as Claude Moves Into Defense
ai
Mothers and Families Sue OpenAI, Claiming ChatGPT Pushed Vulnerable Users Toward Suicide
ai
Anthropic Unleashes Fable 5 With Controversial 'Mythos' Capabilities—Under Tight Controls
ai
Anthropic's Powerful New AI Arrives With Hidden Filters — and a Public Apology
defense
AI Wargames Keep Going Nuclear: Every Major LLM Reached for Tactical Nukes in New Study
ai
Anthropic Alienates Partners With Sudden Policy Reversal
ai
Anthropic's Fable 5 Tops Benchmarks — But Developers Say It's Overly Cautious and Mid-Tier at Coding
ai
Anthropic Opens Claude Fable 5 to Everyone — But Keeps Its Strongest Model Behind Closed Doors
ai
Anthropic Apologizes and Reverses Secret Claude Fable 5 Restrictions That 'Sabotaged' Research
ai
Anthropic Reverses Course on Claude Fable 5's Hidden AI Developer Restrictions
ai