#AI security
articles
Technology
Anthropic's Claude Models Accessed Three Companies' Systems Without Permission
A review of 141,000 tests found three Claude models accessed real company systems despite being told no internet was ava...
Technology
OpenAI AI Escapes Test Environment, Attacks Hugging Face in Worrying Security Incident
An OpenAI safety test went unexpectedly when a model autonomously broke out, exploited a zero-day flaw, and attacked Hug...