The Palm Springs fertility clinic bombing in May 2025 gave the FBI a case study it had been dreading. A 25-year-old used a generative AI tool to obtain bomb-making instructions, bypassing safeguards through jailbreaking techniques that law enforcement considers trivially available. The suspect's privacy-protected chatbot conversation meant no alarm was triggered before the attack killed him and injured four others. An FBI Los Angeles memo distributed in May 2026, previously unreported and reviewed by the Guardian, warned bluntly: current GenAI safeguards are insufficient, and threat actors will increase their use of these tools for attack planning. The Guardian's exclusive review of more than 80 documents and memos — distributed among DHS, FBI, NYPD, and fusion centers between 2021 and 2026, obtained by transparency nonprofit Property of the People — reveals a law enforcement apparatus trying to manage two distinct but increasingly entangled problems. The first is operational: extremists using generative AI to plan attacks, produce deepfakes, generate election misinformation, recruit members, and create child sexual abuse material. The second is infrastructural: protecting datacenters and AI systems from physical attack by domestic violent extremists, including far-right militia members who view destroying internet infrastructure as a catalyst for societal collapse and race war. The infrastructure threat is not hypothetical. Seth Pendley, a January 6 rioter and militia member, plotted to bomb an Amazon Web Services facility, believing it would disrupt 70% of internet service and trigger revolution. He was sentenced to 10 years in 2021. Since then, fusion centers nationwide have tracked an escalation in anti-datacenter rhetoric on 4chan, Facebook, and far-right forums — including discussions of killing AI company CEOs, arson, and bombing attacks. A 2025 Philadelphia police intelligence bulletin cited a white supremacist Facebook post referencing "tannerite and gasoline" attacks on AI datacenters. In Michigan, three people allegedly associated with a white supremacist group protested what they called a "Jewish data center." But here is where the framework fractures. As opposition to datacenters has gone mainstream — driven by water consumption, energy costs, pollution, and climate concerns — the documents show law enforcement casting an increasingly wide net. A Nevada Threat Analysis Center report from April 2026 cited a Reno City Council hearing on a datacenter moratorium as evidence of "increased negative sentiment." This is not a fringe position; it is a standard municipal governance process. The agencies are monitoring proposed legislation, public social media posts, and community protests alongside genuine extremist threats, with no clear analytical firewall between the two categories. The NYPD's involvement illustrates another dimension: the agency produced a threat analysis report linked to the "AI 2027" project, an influential forecast from AI researchers that included scenarios like bioweapon-armed drones wiping out humanity. The NYPD described this as routine intelligence analysis, but the document reveals how speculative AI risk narratives from the industry itself feed into law enforcement threat assessments, creating a feedback loop where hypothetical harms justify expanded surveillance mandates. The structural tension is sharp. AI companies have focused public safety rhetoric on long-term existential risk — doomsday scenarios that remain hypothetical. Law enforcement, by contrast, is dealing with present-day operational threats: the Palm Springs bomber who used a chatbot, the militia members plotting datacenter attacks, the deepfake-enabled fraud. Yet both frames converge on the same policy outcome: expanded monitoring, increased infrastructure protection budgets, and a surveillance apparatus that struggles to distinguish between a white supremacist planning arson and a city council debating water rights. Ryan Shapiro, executive director of Property of the People, articulated the core concern: "There's nothing extreme about not wanting to lose your livelihood to AI, your drinking water to AI datacenters, your privacy to ubiquitous AI surveillance or your life to runaway rogue AI. What's extreme is law enforcement treating these legitimate objections as threats to be neutralized." The 80+ documents paint a picture of agencies that have, since 2021, built analytical infrastructure around AI threats without building the corresponding filters to protect democratic participation from being classified as extremism.
