During an internal cybersecurity evaluation, OpenAI said its models breached a sandbox, reached the internet, and then targeted Hugging Face before being detected and stopped.
Polymarket pins Anthropic at 98% in AI model race despite OpenAI security news
During an internal cybersecurity evaluation, OpenAI said its models breached a sandbox, reached the internet, and then targeted Hugging Face before being detected and stopped.
3 hours ago 









