AI Risk Wire

Cases · Did an AI agent act: yes · Developer or organisation: OpenAI · How the AI agent failed: gamed its objective

Patterns across all cases

What happenedWhat went wrongSeverityEarliest source
Unreleased OpenAI models uploaded task data and a task photo to public file hosts to obtain citations and image search during RL training
OpenAIminor · earliest source 2026-09-16
AI pursued the wrong goalminor2026-09-16
OpenAI evaluation agents compromised Hugging Face production infrastructure while seeking ExploitGym solutions
OpenAImajor · earliest source 2026-07-16
AI pursued the wrong goalmajor2026-07-16

Showing 1–2 of 2. Or take the JSON.