GregoAI earns $250,000 bounty through autonomous exploit discovery
Tech GregoAI earns $250,000 bounty through autonomous exploit discovery A security-focused AI just earned a quarter-million dollars for finding a bug that human auditors missed. No human designed the exploit. No human guided the analysis. The machine did it alone. Grego AI, a startup founded in 2024, announced that its multi-agent system autonomously identified a critical vulnerability in a major blockchain protocol. The flaw could have enabled a $27.7M theft. The affected project responded by awarding a $250,000 bug bounty, which Grego AI says is the largest ever paid for a vulnerability discovered solely by artificial intelligence without human intervention in exploit design. How the system actually works Grego AI calls its approach “Deep Invariant Analysis.” The system ingests a protocols complete codebase, builds dependency maps across the entire architecture, then deploys sandboxed agents that synthesize and test potential exploits. The agents analyze more than seven layers of dependencies, hunting for attack paths that traditional auditing methods might overlook. The sandbox element is critical. Rather than probing live protocols and risking actual damage, the system creates isolated environments where it can attempt exploits safely. When an agent finds something promising, it generates a proof-of-concept exploit to verify the vulnerability is real and quantifiable. A track record of