When it comes to artificial intelligence, the United States has both an extraordinary opportunity and a serious problem.
The opportunity is offered by American technology dominance at the forefront of AI research and development. Maintaining this competitive advantage will allow the public to benefit most from innovations in medicine, information access, economic productivity, and growth that AI promises to deliver.
Continued American AI dominance would also benefit the world by ensuring that the industry is centered in free-market innovation under the rule of law and not subject to the hostile whims of the Chinese Communist Party.
The problem? Beijing knows that whoever dominates AI is likely to dominate the 21st century overall, and every other nation knows it as well. As a result, legions of foreign spies are deployed to steal American AI innovations.
This threat has been exemplified by federal charges filed last week against OpenAI engineer and Saudi citizen Abdulrahman Alfozan. He was caught after attentive OpenAI security personnel discovered a covert recording device hidden in an executive conference room. Alfozan had already booked a one-way flight to Saudi Arabia when he was arrested.
There are two important conclusions to draw from this.
First, while Alfozan has not been directly accused of working as a spy for the Saudi GID intelligence service, that agency prioritizes technology espionage. The GID also adopts an approach similar to that of other U.S. allies, such as Israel and France, serving as an allied partner while also spying on U.S. interests. Saudi state involvement is unconfirmed but nonetheless likely.
Second, it was OpenAI’s security team, not the FBI, that discovered Alfozan’s antics. That speaks to a deeper problem: American counterintelligence (spy hunting) defenses are grossly inadequate. This is so even at the best of times.
But due to FBI Director Kash Patel’s heavy diversion of agents and analysts to investigations involving violent crime, drugs, and immigration offenses, counterintelligence has reached a crisis point.
The Trump administration has done excellent work reversing the Biden administration’s open-door policy toward Chinese students and researchers working in areas sensitive to national security. Far too many of these visitors are willing or cajoled agents of the vast Chinese intelligence apparatus. Secretary of State Marco Rubio’s revocation of thousands of Chinese visas has put a dent in this silent spying campaign.
But that’s only half of the story. As NPR reported earlier this month, FBI counterintelligence squads have been cut dramatically since Trump’s return to office. Some squads are now operating at only half their usual staffing levels. The Washington Examiner independently understands that NPR’s reporting is accurate.
While FBI counterterrorism investigations remain well-resourced, agents must triage heavily to decide which espionage threats to pursue and which to set aside. Because counterintelligence often involves investigating people trained to evade surveillance, each investigation can require significant personnel commitments.
The primary result of this diminished capability is that Chinese spies have a greater chance to steal, and Russian spies have a greater chance to intimidate and assassinate.
The immense economic, social, technological, and political value of AI means every nation that spies on the U.S. views this industry as a top target. For every Alfozan who is caught, 50 others go uncaught.
This threat demands far greater urgency from the Trump administration. A successful espionage campaign targeting U.S. AI would allow a foreign adversary to seize the results of tens or hundreds of billions of dollars in American research and development efforts and use them to develop its own models. We’ve already seen this in China’s distillation of U.S. AI models.
If spies can gain access to nonpublic AI research and models, they’ll threaten the economic viability of America’s AI industry. They’ll also threaten to leapfrog U.S. developments or provide comparable AI systems at far lower cost.
More ominous AI espionage threats include an adversary stealing cutting-edge capabilities and weaponizing them against the U.S. This might take the form of AI-enabled military capabilities or the scouring of software for unidentified flaws.
FOR KATIE ZACHARIA, IT’S ONLY UP FROM HERE
Imagine Chinese AI drone swarms that can confidently detect and evade U.S. military defense systems on Day 1 of a war. Imagine crippling cyberattacks that empty bank accounts, shut down hospitals and transport networks, or otherwise hold our most sensitive information for ransom. Imagine a world in which Xi Jinping or Vladimir Putin gets to set what AI does and does not do.
Patel and the Trump administration need to focus on this. Counterintelligence work might not generate media-friendly headlines and community gratitude of the sort that comes when the public sees gangbangers locked up. But considering the stakes in the race for AI supremacy, counterintelligence investigations must be considered as important as those involving counterterrorism.