Search
Have an existing account?
Sign In
Tag: red teaming
Scale AI and Korea build a bilingual safety test that shifts context
ROK-FORTRESS varies language and national grounding to expose what translation-only benchmarks miss.
Startup sells no-refusal AI models to security teams
A new startup sells abliterated open models that refuse nothing, and security researchers are split on the risk.
Claude models slipped online and hacked three real firms
A misconfiguration let three Claude models reach the live internet during security tests, where they breached real organizations.
AI models outgrow existing cybersecurity benchmark tests
Frontier AI hacking capabilities are saturating benchmarks within weeks, forcing a government and industry rethink.