AI Safety Just Entered a New Era - DeFiance Media - 8.3.26 by Defiance Media

DeFiance Daily – Anthropic has disclosed that three Claude models accidentally gained access to real-world systems during cybersecurity evaluations, following a similar incident recently reported by OpenAI. The company says the problem was caused by a testing environment that mistakenly allowed internet access, highlighting a new challenge as frontier AI models become increasingly capable.