In brief: As dwell time approaches zero, a paradigm shift from detection and prevention toward preemptive resilience with recovery as a design principle becomes necessary.
Frontier AI models like Anthropic’s Claude now identify critical vulnerabilities faster than organizations can patch them. This renders the classical detection model ineffective at preventing damage, because detection often merely documents what has already occurred.
The security principle “Detection buys time” was long based on the assumption that attackers need time to orient themselves in environments, escalate privileges, and execute attacks. This window was the defenders’ buffer. With modern frontier AI models, this dynamic shifts fundamentally. When Anthropic announced in April that the then-unpublished Claude model had identified thousands of critical vulnerabilities across all major browsers and operating systems, a trend became visible: vulnerabilities are being discovered, assessed, and converted into attack paths at machine speed.
The acceleration structurally exacerbates an existing problem. According to available data, more than 88 percent of known vulnerabilities in large enterprises remain unpatched six months after disclosure. At human attack speed, this was manageable through prioritization. However, when AI systems quickly translate new gaps into practical exploits, every open window becomes an invitation. Dwell time—the span until detection and damage—is shrinking from hours or days to minutes or seconds according to Gartner. In this scenario, detection often merely describes what has already happened instead of preventing it.
Source: www.it-daily.net · Published 1 July 2026
Lumi AI News — AI-assisted curation pursuant to Article 50 EU AI Act. Paraphrasing and classification via Lumi News Pipeline v1.7.2.