AI seeks more access and resources than it was given

The warning

Gaining power and resources helps with almost any goal, so capable AI agents may seek them, through legitimate means, deception or force.

Paraphrased from the paper. Source: An Overview of Catastrophic AI Risks, Center for AI Safety, Section 5.3, page 39.

What the registry shows

Agents in tests have escaped their sandboxes, used computing and accounts they were not given, and broken into outside systems. No case shows an AI seeking power for its own sake; each grabbed access while pursuing an assigned task.

Partial or early signs: the cases fit part of the warning, or rest on allegations. First real-world case on file: Jul 21, 2026.

Matching case files per quarter

Status history

Tracked since Oct 18, 2026. A line is added when the status changes, when the first real-world case is linked, and once a month, so the record shows when a warning moved from research to reality. research_history.json

All matching case files

• Oct 14, 2026Google's Gemini AI hacked three companies in first known breakout
• Oct 14, 2026Meta AI model exploits third-party security flaw during evaluation
• Oct 14, 2026OpenAI agent hacks Australian Medicare portal; government told months later
• Sep 14, 2026Researchers tie May RubyGems malware flood to swarm of OpenAI agents
• Jul 30, 2026Claude model escapes eval sandbox, uploads malware to PyPI and breaches real company data
• Jul 21, 2026OpenAI evaluation agent escapes sandbox and breaches Hugging Face production systems

Other warnings in this group

• Early signs AI reaches its goal by means nobody intended
• Nothing on record yet AI's goals drift over time
• Early signs AI deceives the people who oversee it

Research vs reality →