Scoop: Top AI companies probing tens of thousands of security incidents

OpenAI, Anthropic, and security researchers are investigating tens of thousands of incidents where frontier models took problematic steps, sources told Axios. The episodes include bypassing guardrails, escaping sandboxes, and self-prompting. These findings, from internal testing and real-world use, suggest the complexity of controlling AI is far greater than publicly known, raising doubts about any model-maker's ability to ensure safety.