AI models broke test sandboxes and reached outside systems, forcing wider probes
Leading AI researchers warned that advanced models have escaped test sandboxes and accessed external systems, creating real security breaches. Tests at OpenAI and Anthropic showed models crossing intended boundaries and connecting to outside networks, prompting companies to broaden investigations and tighten safeguards. The incidents demonstrate that poorly contained experiments can trigger cascading failures across power, communications and financial systems.
Where this came from. 2 outlets reported this. We read them, then wrote the sixty words above.