FBI warns AI chatbots are becoming planning tools for terrorists

A 25-year-old man detonated a car bomb outside a Palm Springs fertility clinic in May 2025, killing himself and injuring four other people, and the FBI says an AI chatbot helped him get there. The suspect used an unnamed generative AI tool that failed to flag his requests for bomb-making instructions, according to a previously unreported law enforcement document reviewed by the Guardian. The memo also states that the privacy of his conversation with the chatbot meant police were never alerted that someone was researching how to build a bomb.

The case is the clearest public signal yet of a threat US authorities have feared for years: bad actors using generative AI (systems that produce text, images or instructions from user prompts) to plan and execute terror plots. From the FBI down to local police departments, US law enforcement is now warning about two intertwined dangers — terrorists who use chatbots as planning tools, and extremists who target the AI industry itself.

What did the Palm Springs memo reveal?

The memo, described by the Guardian, contains a direct and troubling finding. The suspect's generative AI tool did not detect his request for bomb-making instructions. The conversation stayed private. That privacy, the memo states, hindered any alert to police that a user was seeking bomb-making information.

The blast killed the suspect and injured four people. For much of the public, the bombing registered as a warning about threats to reproductive care facilities. For the FBI, it confirmed a longer-standing concern about AI-assisted attack planning.

The document does not name the AI tool involved. That omission matters. It leaves open whether the failure stemmed from weak safety filters, the inherent difficulty of detecting obfuscated harmful queries, or the encrypted, private nature of chatbot conversations themselves.

How do authorities expect extremists to use AI?

The law enforcement document lays out several areas of concern, each of which maps to a distinct policy problem for AI developers and police alike:

• Attack planning and recruitment. How extremists could use AI to devise attacks and recruit new members.
• Synthetic harmful content. How AI could be used to produce and disseminate everything from deepfakes (fabricated but realistic media) to election misinformation and child sexual abuse material (CSAM).
• Circumventing detection. How people seeking information on how to make weaponry or bombs could use AI to evade police detection — the exact pattern seen in Palm Springs.
• Policing with AI itself. How officers could use AI tools in policing, for tasks like social media monitoring, while managing public concern over the deployment of those same tools.
• Anti-AI extremism. Monitoring social media for opposition to AI and for potential extremist attacks on datacenters, the physical facilities that house the computing infrastructure behind AI systems.

That last item signals a shift in how law enforcement frames the AI threat. The technology is simultaneously a tool and a target. Police must police the misuse of systems that some violent actors want to physically destroy.

Why does this matter for AI policy?

The stakes cut across the AI industry, public safety and civil liberties.

For AI developers, the memo sharpens pressure on safety filters. A chatbot that answers bomb-making queries without triggering an alert is a product failure with lethal consequences. Companies already face regulatory and legislative scrutiny over harmful outputs; a documented case connecting a chatbot to a real bombing gives regulators a concrete anchor.

For lawmakers and platforms, the privacy question is harder. End-to-end privacy in AI conversations is a feature for ordinary users and a blind spot for investigators. The memo explicitly identifies that privacy as an obstacle to alerting police. Any move toward monitoring or reporting suspicious AI conversations would raise significant civil liberties concerns — and would need to reconcile with how tech companies already handle similar duties for other content, such as CSAM reporting.

For police departments, the memo reflects an institutional dilemma. Agencies want to deploy AI for social media monitoring and other investigative tasks, but the document acknowledges they must manage public concern over exactly those uses. Public trust and operational capability are in tension, and the memo puts that tension in writing.

What does the anti-AI extremism threat look like?

The document instructs attention to a threat vector that has received far less public discussion than AI misuse: violent opposition to AI itself.

Law enforcement is monitoring social media for opposition to AI and for potential extremist attacks on datacenters. These facilities — large buildings packed with servers running AI workloads — are physically vulnerable, geographically fixed and increasingly central to the economy. An ideology that frames AI as an existential threat could, in the FBI's framing, motivate attacks on that infrastructure.

The dual framing — AI as an instrument of terrorism and as a target of terrorism — places the industry in an unusual position. Companies must harden their products against misuse while also considering the physical security of the infrastructure that runs them.

What comes next?

The Palm Springs case is unlikely to remain an outlier in FBI threat assessments. The memo reviewed by the Guardian shows the bureau has already moved from abstract concern to documented incident analysis, and the list of concerns it enumerates — recruitment, deepfakes, election misinformation, CSAM, weapons instruction, datacenter attacks — reads as a working agenda for both AI safety teams and counterterrorism investigators in the months ahead.