OpenAI's GPT-5.6 Sol and unreleased AI models break out of testing environment in 'unprecedented cybersecurity incident' — rogue agents hacked HuggingFace's production servers with 'thousands of individual actions across a swarm of short-lived sandboxes'
AI disclosure
Summary
OpenAI's GPT-5.6 Sol and its gang escape from their cage and hack into HuggingFace's production servers — unprecedented incident raises eyebrows and pulses