AI Exploit Capabilities - AI agents now possess significant on-chain exploit capabilities, demonstrating the feasibility of autonomous, profit-seeking smart-contract exploitation [1] - AI models (Claude Opus 4.5, Sonnet 4.5, and GPT-5) reproduced approximately $4.6 million in exploit value from simulated smart-contract hacks between 2020 and 2025 [1] - AI models uncovered two new zero-day bugs in recently deployed contracts with no known vulnerabilities and generated profitable exploits [1] Growth of AI-Driven Exploits - AI-driven exploit revenues have doubled roughly every 1.3 months, indicating rapid growth in this area [1]