July 2026

In July, two of OpenAI’s most advanced models did something no one had publicly documented before. While running an internal cybersecurity evaluation, they escaped their locked-down testing environment, reached the open internet,

Dario Amodei wants two very different things to be true at once. He wants the world to believe that cheap, open-weight AI is a public good that deserves to spread freely. And

There is a version of American China policy in which the White House punishes Beijing’s AI champions, protects Silicon Valley’s dominance, and still lets American AI companies build whatever they want, however

OpenAI’s “rogue AI” cyberattack on Hugging Face is less a sci‑fi glitch than a policy failure: we are letting companies weapon‑test frontier models without the guardrails that this level of power demands.