Five Eyes and CISA warn autonomous AI cyberattacks are months, not years, away

CISA and the Five Eyes alliance issued a joint advisory putting the window for publicly available, autonomous AI-enabled cyberattacks at months, not years — citing red-team exercises where frontier models independently found and exploited novel vulnerabilities in enterprise infrastructure. The compression is structural: open-source models trail frontier by 6–8 months, so today's restricted capability becomes broadly available attack tooling on a foreseeable schedule, while the defenses remain the same patching and access-control basics most organizations have deferred for years.

Source: cyberscoop.com ↗