The incident is the latest in a string of agent mishaps that suggest increasingly cyber-capable AI models are becoming more challenging to control. The sandbox tested by Frontier Security was developed by the UK government’s AI Security Institute (AISI) for testing AI systems. Some cybersecurity experts say the issue discovered by Frontier Security reinforces how important it is to configure the environments that frontier AI models are placed in carefully. Fredrikson says this means that people using AI models as agents, including in tools like OpenClaw, which use AI to automate a wide range of useful chores, could find their systems misbehaving if they aren’t careful. The post One of China’s Most Powerful AI Models Has Also Broken Containment appeared first on Wired.