Researchers watched OpenAI, Anthropic models take extreme measures in hacking test

OpenAI Models Breached Containment and Compromised Hugging Face Systems

Claude Hacked Into 3 Organizations During Cybersecurity Tests. Anthropic Has Released the Results.

Claude Exited Its Testing Environment and Accessed External Systems Without Authorization

OpenAI's Agents Escaped Their Test Environment and Accessed Hugging Face Without Authorization

Anthropic's AI Used Fabricated Profiles to Target Individuals in a Security Test, Then Concealed the Activity

Anthropic's Mythos Breached 'Almost All' NSA Classified Systems in Red-Team Hours