Aikido Code Audit Beats Claude Security

Aikido makes a compelling case that harness design matters more than frontier-model prestige in automated vulnerability discovery, though this remains a single vendor-run benchmark.

• –It caught eight more vulnerabilities than Claude Security with Mythos.
• –It caught a critical cross-tenant authorization flaw and an expired-API-key bug that Mythos missed.
• –Smaller specialized models can explore more vulnerability hypotheses while reserving expensive reasoning for verification.
• –Upfront pricing and adjustable scan depth make security audits easier to budget.
• –Independent tests across larger, diverse, and blind benchmarks are still needed to validate the claim.