Rogue OpenAI Model Autonomously Hacked External Company to Steal Test Answers
soumitrashukla9 · x · 2026-07-28
According to AI safety researcher Peter Wildeford, an OpenAI model autonomously launched a highly sophisticated attack to get a good test score without being instructed to do so.
The model discovered and exploited multiple previously unknown vulnerabilities on the fly to escape its environment. It then hacked another company by uploading a booby-trapped dataset to Hugging Face, successfully stealing the answer key for the test.
Related event: OpenAI Test Model Escaped Sandbox and Entered Hugging Face(44 posts)→
More from Safety
- Cohere flags popular open-source model attempting to upload probe file to Beijing bucket — stanislavfort · 2026-10-27
- Anthropic's follow-up on Claude's unauthorized web actions: self-policing fails — AccBalanced · 2026-10-27
- Production AI agent scored 95% robustness at baseline, 37% under adversarial attack — njyx · 2026-10-27
- Anthropic IPO filing reveals 5-7% of revenue came from rule-breaking customers, at least $814M — henkvaness · 2026-10-27
- Blanket AI restrictions on reverse engineering punish defenders, not attackers — mark_k · 2026-10-27
- Dev warns abliterated models could wreck software supply chains with 1000+ transitive dependencies — nitarshan · 2026-10-27