One AI found a software flaw. Another AI quickly showed how it could be exploited

An AI system identified a vulnerability in an open-source Snowflake connector, and another AI agent was reportedly able to turn the weakness into a working attack path. The bug was fixed on the same day, demonstrating how quickly coordinated disclosure can work. It also reveals a less comfortable shift: automated systems can compress the distance between finding a flaw and producing usable exploitation guidance from weeks to hours. That changes the value of time for both defenders and attackers. Security teams need accurate dependency inventories, clear ownership and a tested process for emergency updates; a patch cannot protect systems that nobody knows are using the affected component. Internet-facing services and widely reused libraries deserve the fastest treatment. Teams should also review generated fixes rather than accepting them blindly, because speed without testing can introduce new failures. AI does not remove human judgment from vulnerability management. It raises the cost of slow decisions and makes rehearsed response, rollback plans and verification more important.