--
[AI Discovers $4.6 Million Smart Contract Vulnerability in Simulated Testing] Anthropic identified multiple smart contract vulnerabilities through AI agents during advanced red team testing, with a potential exploit value of $4.6 million. The testing process included contract analysis, command-line tool configuration, network reconstruction, exploit generation and validation, and introduced a new benchmark for smart contract security assessment. This research was jointly conducted by Anthropic, the MATS project, and the Fellows program.