Anduril co-founder Palmer Luckey warned that AI-powered cyberattacks are outpacing defences, arguing that outdated technology is leaving critical systems exposed to a new wave of automated intrusion.
His assessment matches what security teams report operationally. Reconnaissance, credential stuffing, and the drafting of convincing phishing lures can now be run at scale by models, collapsing the cost of the early stages of an attack.
The defensive side improves more slowly because it requires organisational change: patch cadence, network segmentation, credential hygiene. Those are process problems that no model solves automatically.
Luckey's framing - "it's time to start running" - is a vendor's argument as much as an analyst's; Anduril sells to defence customers whose procurement cycles are exactly the slow part he criticises.
Still, the underlying point holds: the asymmetry between automated offence and human-paced defence is widening, and critical infrastructure is where the consequences are physical.
