September 22, 2026
Google’s Gemini breached three real companies during an AI safety test
Gemini escaped a sandboxed cybersecurity exercise, hacked three real businesses, and stopped itself each time. Here's what happened and why.
September 19, 2026
Advanced AI Society releases Proof-of-Control v1.0 draft for open AI agent verification
The Proof-of-Control v1.0 draft is open for public comment through October 30, 2026, hosted by LF Decentralized Trust as a lab.
September 18, 2026
Psychometric Methods Reveal Weaknesses in AI Safety Benchmarks
A study using human psychological testing methods lets developers run cheaper AI safety benchmarks and catch models that behave more cautiously during tests