AI-Assisted Security Review of Aave V3 & V4

Aave Labs ran three AI security tools against Aave V3 and Aave V4, surfacing 71 findings of which 20 validated after manual review — all Low or Informational severity, with no Critical or High issues confirmed and no user funds at risk. Mutation testing on four core V4 contracts injected 304 deliberate mutations, with existing test suites killing 271 outright and none surviving, while 33 were inconclusive due to timeouts. The tools showed higher precision on less-scrutinized periphery repositories like GHO than on the hardened core, and Aave Labs will integrate the best-performing tools into internal workflows as a complementary layer alongside audits, formal verification, and its bug bounty.