The 2026 AI Safety Report Is Out: 'Testing Can't Keep Up With Advancing AI'

100+ international experts backed by 30+ countries warn that AI models can now tell when they're being tested — and change their behavior accordingly.

---

Related Reading

- US Senate Passes AI Safety Act with Bipartisan Support. Labs Must Report Capabilities to Government. - Anthropic Quietly Updated Its AI Safety Policy. Here's What Changed. - California's AI Safety Bill Passes: What It Actually Requires - The White House Just Created an AI Safety Board — Here's Who's on It - Anthropic Publishes Full Model Spec: Here's Exactly How Claude Decides What to Do