Former National Security Agency (NSA) Director Keith Alexander said on Oct. 29 that artificial intelligence (AI) and unmanned systems will reshape military operations, but warned that their effectiveness depends on reliable intelligence, human judgment, and the ability to act quickly.
Speaking at Cloudera’s Evolve 2026 Government Summit in Washington, D.C., Alexander – who also served as the founding commander of U.S. Cyber Command – called for faster technology development, greater collaboration between military personnel and innovators, and stronger safeguards against inaccurate data.
From 16 hours to 1 minute
Alexander pointed to his experience in Iraq to illustrate the operational consequences of intelligence delivery delays.
He recalled a situation in which intercepted intelligence was taking 16 hours to reach personnel who needed it to locate targets. Bringing operational personnel and technology experts together to redesign the process reduced that time to approximately one minute.
“We got the people to come out, and we got the tech guys. We built the system,” Alexander said. “They went from taking 16 hours down to one minute to locate a bad guy.”
AI and unmanned systems to reshape warfare
Looking ahead, Alexander identified AI and unmanned systems as technologies that will fundamentally change military operations.
He pointed to drone warfare as an example of the need to collect intelligence, process large volumes of data, and deliver targeting information quickly. That will require cloud infrastructure, systems that can process and distribute data within milliseconds, and affordable targeting capabilities, he said.
Alexander advocated for giving technology experts clear operational objectives and the freedom to develop solutions – an tactic he described as the “Star Trek approach.”
He also emphasized testing emerging capabilities in realistic military environments, including the Army’s National Training Center, before expanding their use.
“AI gives us the ability to actually iterate through that process in ways that we couldn’t in the past,” Alexander said.
Such capabilities could help the military test approaches, evaluate results, and refine technologies more quickly.
Human oversight remains critical
Alexander cautioned that AI’s effectiveness depends on the quality of the information it receives and independent human evaluation.
“Bad data in, bad data out,” he said.
Inaccurate or misleading information can produce flawed conclusions, which means trained personnel are essential to identifying errors and questioning AI-generated results.
“And the best one for detecting that, the best thing, is people,” Alexander said. “Trained people who know the difference between right and wrong, and can ask that. Let me check that.”
Alexander distinguished between challenges with AI models and the persistent challenge of unreliable source data.
“But we’ll build systems that eventually get rid of hallucinations and things like that and the drift. But you’re always going to have that [challenge] in bad data,” he said.