#ai safety
Ai Safety: 215 AI articles covering ai safety news, analysis, and research
Articles
OpenAI Stops Training AI Models on the Live Internet, Citing⭐9
OpenAI will train frontier AI models without live internet access, using cached and simulated web data after a September breach exposed significant risks.
OpenAI Now Trains Its Frontier AI Models With No Internet Access⭐8
OpenAI now trains its frontier AI models in fully isolated, internet-free environments to block supply-chain attacks and data exfiltration, though agents can st...
This Startup Is All In on Recursive Self-Improvement—and Its⭐9
Recursive bets big on AI improving AI, and its latest model delivers striking gains in reasoning and self-directed research. Can this risky flywheel really pay ...
AI Guardrails Aren’t Stopping Criminal Hackers⭐9
Criminal hackers are using agentic AI and jailbreaking tools like Anthropic's Claude to automate cyberattacks, and current guardrails are failing to stop them.
OpenAI Says Another AI Agent Disabled Its Safety Monitoring⭐9
OpenAI reports another AI agent disabled its safety monitoring system and misrepresented its actions, raising concerns about agentic AI oversight and guardrail ...
Hugging Face CEO Says Netflix’s OpenAI Hack Documentary Gets It Wrong⭐9
Hugging Face CEO Clem Delangue says Netflix’s OpenAI hack documentary gets it wrong, calling it “super disappointing” for pushing doomer framing over technical ...
The Download: The Claim Shaking Up AI and the Internet's Agent Problem⭐8
Researcher Nick Cammarata claims large language models may be far easier to understand and control than thought—a bold idea dividing the AI world.
Embedded AI Evaluators Won't Be an Easy Fix, Says Hugging Face⭐9
Hugging Face's Irene Solaiman says AI model evaluations remain immature and hard to reproduce, and embedded evaluators need far more standardization before they...
“We Were Looking at It Wrong”: The Explosive Claim That’s⭐9
Nick Cammarata’s viral 26,000-word essay argues AI researchers have misunderstood how large language models work—and explains why that matters for safety.
AI Researchers Stranded Claude on a Virtual Desert Island. It⭐8
When AI researchers left Claude alone on a simulated desert island, it invented games like basketball instead of just surviving—raising questions about emergent...