AuditAgent
AuditAgent on EVMBench: 40 Repositories, 120 Vulnerabilities, No Repos SkippedAuditAgent detected 80 of 120 high-severity vulnerabilities on EVMBench. 40 repositories, run in order, no repos skipped. Full results and methodology.
AuditAgent Back to Blog AuditAgent on EVMBench: 40 Repositories, 120 Vulnerabilities, No Repos Skipped April 7, 2026 Most AI security tools do not publish standardized benchmark results. Standardized benchmarks produce numbers that can be compared, and comparisons require confidence in methodology. EVMBench is one of the first standardized benchmark for AI vulnerability detection on smart contracts, built by OpenAI and Paradigm. We ran AuditAgent across all 40 repositories. Here is what it shows, how it was run, and what the numbers mean. What Is EVMBench and What Does It Measure EVMBench cove
saved by
related reading
- EVMBench Leaderboard — AI Smart Contract Auditorstestmachine.ai
- evmbench.pdfcdn.openai.com
- Demystifying evals for AI agents \ Anthropicanthropic.com
- AuditBenchalignment.anthropic.com
- How Figma stays ahead of vulnerabilities with agentsfigma.com
- Center for Responsible, Decentralized Intelligence at Berkeleyrdi.berkeley.edu
- Incident Report: unsanctioned agent behaviour during cyber testing | AISI Workaisi.gov.uk
- Building and evaluating alignment auditing agents — AI Alignment Forumalignmentforum.org
- Senior SWE-Benchsenior-swe-bench.snorkel.ai
- BenchmarkList: Track the Frontier of AI Capabilitiesbenchmarklist.com
- [2602.22755] AuditBench: Evaluating Alignment Auditing Techniques on Models with Hidden Behaviorsarxiv.org
- Vulnerability Research Is Cooked - Quarrelsomesockpuppet.org