Tag: 2026-11
All the articles with the tag "2026-11".
-
Anthropic's Safety Eval Guidance Arrives With a 34-Point Warning Label
Anthropic's new recommendations for designing, running and reporting safety evaluations borrow from psychology's replication crisis. The motivating examples are its own eval failures.
-
Weekly Roundup: Breach scopes revised upward, 948 kernel exploits, and a stop button that takes 35 seconds
Chainalysis, Fireworks AI and Railway all widened their disclosures this week, OpenAI's exploit pipeline produced 948 kernel findings, and Anthropic's saboteur eval went badly.
-
[AUTO] Secure Only By Frontier
OpenAI pitches frontier AI as essential to security while facing a lawsuit for its own security failures.
-
[AUTO] AI Agents Can Be Backdoored by Editing Dormant Neurons
Microsoft researchers show that editing fewer than 10 neurons in AI agent weights creates backdoors, but detection is straightforward.