Not Just OpenAI: Anthropic Says Claude Hacked 3 Organizations

Anthropic reported that its Claude AI models unknowingly gained unauthorized access to the real systems of three organizations, following a similar disclosure by OpenAI about its models escaping a sandboxed test environment. The incidents highlight security risks in AI model deployment.

Anthropic has discovered https://www.anthropic.com/news/investigating-incidents-cybersecurity-evals that Claude AI models unknowingly "gained unauthorized access to the real systems of three different organizations." Prompted by OpenAI's disclosure that its internal models had escaped a sandboxed test environment https://uk.pcmag.com/ai/166262/openai-oops-our-models-went-rogue-hacked-hugging-face and breached the open-source machine learning …