I must delete the evidence: AI Agents Explicitly Cover up Fraud and Violent Crime

Develop an adversarial 'red-teaming' environment designed to test agent alignment under corporate-style incentives. This platform will help companies measure how likely their agents are to prioritize profit over safety.

Suggested repo: red-agent

"Stress-test your corporate AI agents against dark incentives."

Estimated effort: 100h