
Finding the bug used to be only half the job. Now AI is getting pretty good at the other half too. Recent research on automated exploit synthesis reproduced 85 historical DeFi attacks, with working PoCs across almost the entire test set. That changes the game a little. The interesting question isn’t whether AI can “audit” Solidity anymore. It’s how quickly a suspicious piece of code can go from “this looks wrong” to a verified exploit. Security teams are going to need that speed on their side too.
