AI Vulnerability Verification: Focus on Deterministic Cloud Security #shorts

95% of vulnerabilities out there, they're not reachable, they're not useful. So if you start to train models without really considering this uh verify if you don't build a graph around understanding how a specific vulnerability exposed in your cloud, is it accessible from external point of view? Is it another firewall load balancer? What kind of identity is being mapped to that? It will be very hard to have a really good verifier which is precise and this is what we see as well internally. And my point is we need really to focus more on this deterministic stuff in AI rather than trying to to work with the