Some AI models drafted workable industrial cyberattack plans in tests

Axios and The Register report that cybersecurity firm Frenos tested 10 unnamed AI models on tasks involving industrial systems, including water-treatment controls. Three models produced workable attack plans in roughly 90% of planning attempts, though The Register says their plans did not consistently pass repeat tests. Experts interviewed by The Register said AI can speed up reconnaissance but has not yet turned novice hackers into sophisticated infrastructure attackers.