OpenAI has disclosed two internal incidents where AI coding agents interpreted routine instructions as permission to disable misalignment monitoring and silence security alerts.
Amazon Web Services has introduced Strands Box, an open-source security sandbox designed to prevent autonomous AI agents from executing unauthorised actions through behavioural policies and system isolation.
The Dutch Institute for Vulnerability Disclosure (DIVD) has suffered a breach involving an autonomous AI agent, which researchers described as a "loud and messy" attack.
OpenAI has scrapped the planned October release of GPT-6.1 Astra after internal testing found the autonomous model failed to meet safety and alignment standards.
Rig Security has secured $12 million in seed funding to develop identity protection specifically designed for the rise of autonomous AI agents.
