Palmer Luckey Warns AI-Powered Cyberattacks Are Outpacing Defenses

Palmer Luckey, founder of Oculus and co-founder of Anduril Industries, has warned that AI-powered cyberattacks are moving faster than outdated technology systems can defend against them. He made the remarks at The New York Times DealBook Summit, in an interview with Andrew Ross Sorkin, as TechCrunch reported on 18 November 2026.
Luckey said advanced AI tools are enabling the creation of various exploits, meaning working code that takes advantage of a software flaw. He argued that U.S. adversaries are exploiting the gap between new attack capabilities and defense capabilities, for both strategic and financial reasons. He singled out North Korea as particularly involved.
Vice President J.D. Vance canceled his appearance at the same New York Times conference because of recent tech outages, in order to attend meetings on the matter.
Looking further ahead, Luckey said advantages will accrue to the side most effectively using AI for defense. He told organizations they will have to pick up the pace of securing their defenses, and summed it up in four words: "It's time to start running."
In my view, the most useful part of the warning is its framing of the problem as a race of deployment speed rather than a contest of model capability. Attackers do not need a breakthrough model. They need tools that shorten the path from a discovered vulnerability to a working exploit. Defenders face the harder version of the same problem. Patch pipelines, change-control boards, and legacy systems built for a slower adversary set the tempo of remediation, and those processes do not compress easily.
The practical questions for security teams are concrete. How quickly can a newly disclosed weakness be triaged, tested, and closed across production systems? How much of the estate runs software its vendor no longer updates? Where can AI-assisted code review, fuzzing (automated testing that feeds unexpected inputs to software to provoke crashes), and configuration analysis run before release, so that defenders find flaws first? Luckey's argument implies these are no longer optional modernization projects. They determine how exposed an organization is when the next exploit appears.
The broader context here is that the same capability cuts both ways. Tools that generate exploits can also read a codebase, map its attack surface, and propose fixes at a scale no human review team can match. Procurement cycles, skills, and the willingness to automate remediation will matter as much as the models themselves.
There is also a geopolitical dimension. If adversaries are gaining from the gap for strategic and financial ends, closing it becomes a matter of national resilience, not only enterprise IT. That is why a sitting vice president rearranged his schedule around the outages and the meetings that followed.
Luckey did not offer a timeline, a list of specific tools, or a regulatory proposal in the remarks reported. He offered a clear instruction to move faster. For engineers who spend their days shipping and securing systems, that instruction is the part most likely to change next week's planning meeting.


