London startup CodeWall raises $4M pre-seed to test enterprise security with autonomous agents

The startup claims its agents can detect and verify attack paths, providing enterprises with reproducible proof-of-concept evidence. CodeWall cites a March 2026 test of McKinsey’s Lilli AI platform, where its agent gained read and write access to the production database in under two hours due to an unauthenticated interface and SQL injection flaw. The company disclosed findings to McKinsey and confirmed remediation before publication. However, the vendor’s own research should not be interpreted as evidence of ongoing vulnerabilities or a benchmark against human testers. CodeWall’s funding will support expanding its testing capabilities, hiring, and refining remediation workflows, though it has not yet disclosed independent performance metrics or named customers.