Home/AI Incident Database/Players convince the "Freysa" AI agent to transfer its $47,000 prize pool despite instructions never to send money
Players convince the "Freysa" AI agent to transfer its $47,000 prize pool despite instructions never to send money
In a crypto game an LLM agent guarding a prize pool was told never to approve transfers; a contestant's message redefining the tool's meaning made it send the entire pool.
Contribute
Spot an error, or know a record we should add? Members can propose additions and submit corrections.