NSA official turned founder: AI multiplies attackers’ expertise while defenders are still short on talent

CrowdStrike’s latest read on a hacker points to something uncomfortable: the operator did not look like an elite team or a state program. Instead, one person paired agentic tools with older techniques, picking Korean lenders as targets, steering the software at them, and asking Claude where stolen data might be sold. The machine did most of the heavy lifting. The human supplied the aim.

That pattern isn’t limited to one case. In November 2025, Anthropic said a group it believed with high confidence to be Chinese state-sponsored used Claude Code in an espionage campaign against about 30 organizations. Anthropic estimated the AI handled 80% to 90% of the tactical work. People still chose the targets and gave the final sign-off before reconnaissance turned into exploitation, and the model was not exactly omniscient: it often overstated what it found and sometimes invented data.

That is the real shift. With AI, attackers can stretch one person’s reach across multiple institutions in weeks, work that used to demand a skilled team. But the defensive side still runs into the same old bottleneck: there are not enough people who can sort real attack paths from a flood of machine-made noise.

The source is personal for a reason. Before co-founding Synack in 2013, the author worked in computer network operations at the NSA, where even extraordinary tools could not decide what mattered. People did that. They picked the opening, judged the risk, and improvised when a target behaved in a way no playbook expected. The best operators were not always the ones with the fanciest tools. They were the ones who saw the path nobody else saw.

That is also why the Synack-NetSPI merger, completed in October, matters. Both companies spent years building offensive security teams, and the article argues that AI makes that kind of judgment more valuable, not less. It lowers the cost of attacking, which means defenders who can spot the few important findings become even more important. The scoreboard is not headcount. It is how far one sharp person can reach.