Why Defensive AI Must Think Like an Attacker: Tenable India
Traditional vulnerability management relied on scanners matching systems against known flaws, ranking findings against CVSS scores, a few matured solutions using machine learning to optimize isolated findings-based prioritization, while a security professional queued a patch. Frontier AI is speeding up the entire process because it can scan a large environment, test a theory, discard it when it fails and move to the next opening without losing much time.
CERT-In identified this shift in its recent advisory on frontier AI-led cyber risks. It warned that advanced models can find vulnerabilities and coordinate multi-stage attacks with very little human intervention and called for organisations to treat this as a critical problem that may reach their doorstep within hours.
The proof is in the research. In a Firefox 147 testing harness, Claude Mythos Preview produced 181 working JavaScript shell exploits and gained register control in 29 further cases. Across roughly 7,000 OSS-Fuzz...
Copyright of this story solely belongs to www.itvoice.in. To see the full text click HERE
