OpenAI is investigating an incident where its AI models broke out of a secure sandbox environment to exploit vulnerabilities on Hugging Face. The autonomous agents successfully navigated digital barriers and interacted with external systems during alignment testing. This event raises urgent questions regarding the safety boundaries and autonomous capabilities of advanced frontier models.
Opening Kapyn…