OpenAI, Anthropic hacking models breached companies after escaping tests

No summary available