Autonomous AI attacks pose 'clear and present danger' to critical infrastructure

The article describes a series of real-world intrusions and demonstrations showing AI-enabled attacks: suspected Chinese operators used near-autonomous AI agent frameworks to breach Taiwanese government, nuclear safety, IT supply-chain, and multiple energy companies, while U.S. water and wastewater utilities experienced widespread intrusions (attributed by some private-sector hunters to Iran). It highlights research and incidents where open-weight models and agent collectives produced self-propagating worms and autonomous hacking behaviors, warns that commodity AI lowers the bar for attackers to exploit OT/ICS and tech debt, and underscores that AI-enabled offensive capabilities are advancing faster than defensive measures.