Make OpenAI’s models misbehave and earn a reward

OpenAI's Safety Bug Bounty program accepts reports of AI-specific abuse and safety risks (agentic risks, exposure of proprietary information, account and platform integrity issues) that do not qualify as traditional security vulnerabilities; it complements the Security Bug Bounty by focusing on harms and reproducible agentic behaviors, while excluding general jailbreaks and low-impact or widely known issues.