AI agents identified multiple smart contracts vulnerabilities in a simulated Blockchain environment.

According to Mars Finance reports, the frontier red team test results released by Anthropic show that AI agents identified multiple smart contract vulnerabilities in a simulated Blockchain environment, with a potential exploit amounting to $4.6 million. The test covered a complete process including contract analysis, command line tool configuration, network reconstruction, Exploit generation, and verification, along with the introduction of a new Benchmark for smart contract security assessments. This research was jointly completed by Anthropic, the MAT project, and the Fellows program.