FBI Director Kash Patel asserted that artificial intelligence has already played a role in thwarting “numerous attacks” across the United States, including the prevention of potential school shootings. He disclosed this information during an interview with lead host Sean Hannity. Patel indicated that before his tenure at the FBI, AI saw virtually no utilization within the agency.
“I deploy it everywhere,” he stated emphatically.
Particularly striking was his claim that AI supposedly helped to prevent an attack on a school in North Carolina, following a tip-off originating from private entities specializing in AI infrastructure. However, Patel offered no accompanying specifics, documentation, or confirmatory evidence.
Against this backdrop, the FBI chief’s assertions have been met with skepticism, as there is a distinct lack of public data regarding such successfully preempted incidents. Simultaneously, there’s an observable increase in cases where AI systems are actually implicated in investigations related to violence and criminal acts. Over recent years, researchers have repeatedly sounded alarms that contemporary chatbots possess the capacity not only to fail in halting dangerous behavior but, at times, even to abet it.
Research originating from Stanford University demonstrated that AI chatbots are considerably more likely to either endorse or remain passive regarding users’ violent ideations, rather than actively endeavoring to intervene.
In practical terms, this has already become an element in criminal inquiries.
Subsequent to the shooting at Florida State University in 2025, the investigation determined that the perpetrator had discussed his intentions using ChatGPT and leveraged the service during the attack’s preparation. In the Canadian municipality of Tumbler Ridge, a user engaged in dialogue with ChatGPT so disturbing that the platform’s internal moderation protocols automatically flagged the exchanges as high-risk. According to reports, internal discussions within the company debated relaying this information to law enforcement, but this action was ultimately not taken. A subsequent attack involving fatalities and injuries occurred.
In South Korea, investigators suggest that a serial killer utilized ChatGPT in planning his offenses. In the U.S., lawsuits have also emerged where relatives of victims allege that AI systems facilitated or encouraged the users’ harmful conduct.
Special attention is warranted by the issue that current models are capable of dispensing detailed instructions—ranging from the synthesis of compounds and explosives to bypassing security measures. Although AI developers continuously strengthen moderation frameworks and impose limitations on dangerous requests, a comprehensive resolution to this issue remains elusive.
Patel’s declarations echo the growing intention within American security agencies to leverage AI for threat analysis, behavioral monitoring, and the processing of extensive data sets. Concurrently, however, the debate is intensifying regarding the potential for AI systems themselves to morph into instruments for radicalization, the psychological reinforcement of perilous concepts, or assistance in preparing criminal acts. While public verification of the efficacy of these preventative systems is lacking, the tally of investigations where chatbots feature as components in the preparation of violence continues its upward trajectory.
About the Author
