Blog

Trust

Security, safety, permissions, evaluation, reliability, and what can go wrong.

An AI Agent Found a Live Admin Key in 25 Minutes: The Baseten Token, Explained

Security firm Strix pointed one AI agent at Baseten's public surface. In 25 minutes it found a three-year-old GitHub token with admin rights sitting in a public Docker image.

Who Is Actually Winning the AI War? The One Thing China and Trump Agree On

Anthropic's CEO asked the industry to slow AI down. Within a day China called it fearmongering and Trump called guardrails a sick conspiracy. Two rivals, one refusal.

Is the Internet About to Be Taken Over? Amodei's Pace the Frontier Warning

Anthropic's CEO says an AI agent swarm could take over the entire internet within 6 to 12 months, and asks every lab to slow down. What the post says, and what it commits to.

The Hidden Danger in AI Agents: Anthropic's Threat Report, Explained

Two students ran 13 unsupervised AI agents harvesting military intel. A doxxing site built by one person. Anthropic's report on eight months of Claude misuse, and what it admits it cannot see.

AI Kills Us All by 2030? What the Fight Is Actually About

An ex-Anthropic researcher says AI could kill us all by 2030 and that it is not a stunt. Elon Musk called it a psy op. Here is what the argument is really about.

The AI Model Google Won't Let You Touch: Gemini 3.8 Flash Cyber

Google's Gemini 3.8 Flash Cyber writes 2.6x more correct Chrome security patches than bigger rivals, yet almost no one can use it. Here is who gets in and why.