AI-assisted vulnerability discovery is increasing the pressure on organizations to patch software before attackers exploit newly disclosed flaws. Security specialists told WIRED that this shorter window can force under-resourced teams to deploy updates with less testing, raising the chance that a defensive fix causes an outage.
Recent disruptions at Microsoft, Apple and the US National Security Agency were not a coordinated AI attack. Microsoft cited a flawed infrastructure update and other network failures, while the NSA’s internal communications outage followed problematic patching. The broader concern is pace: more than 100,000 vulnerabilities have been recorded in 2026, compared with about 25,000 in 2022.
Raw discovery counts overstate immediate danger. The US Cybersecurity and Infrastructure Security Agency has added only about 300 flaws to its exploited-vulnerability catalog this year; VulnCheck counts roughly 1,500 known exploited issues. AI can help administrators triage fixes, but researchers stressed that people still need to diagnose causes and validate patches. Years of deferred maintenance make that compressed decision cycle harder to manage safely.