A vulnerability scanner answers one question: which known CVEs are in this thing right now? That's useful and it's also reactive. It only ever tells you about what's already been detected. The question a security team actually loses sleep over is the next one: what am I likely to have that hasn't shown up yet? What's the vulnerability an attacker finds next, in the same component, that my current scan didn't flag?
At RapidFort I built a service to take a swing at that question, an AI-assisted sister-CVE predictor . The idea: given the CVEs a product already has, predict the related ones it's likely to have too. It used ExploitDB as the ground truth, Google's Gemini API as the analysis layer, and shipped as a Kubernetes microservice inside our AI CVE-prediction feature. It landed well with the team and, more importantly, with customers, because it turned a flat list of findings into a head start.…