California fertility clinic bombing showed how generative AI can aid terror plots, FBI warns

An investigation into a bombing outside a fertility clinic in Palm Springs in 2025 shows how U.S. law enforcement agencies are thinking about artificial intelligence (AI) and terrorism.

Investigators say the suspect used a chatbot to ask how to make a bomb. This raised concerns about gaps in the safety rules used by AI platforms and the usual ways authorities try to spot threats before violence happens.

Here's what to know

A 25-year-old man is believed to have set off a car bomb outside the California clinic in May 2025. He died, and four other people were injured.

A law enforcement document that had not been reported before, obtained by the Guardian, said the man had used an unnamed generative AI tool to ask for bomb-making instructions.

The document also said the platform had not detected or flagged the request about violent plans. Because the AI conversation was private, it was harder for authorities to learn about the threat before the attack.

The Guardian reviewed more than 80 federal, state, and local documents from 2021 to 2026. They showed that law enforcement agencies were becoming more concerned about possible links between AI and extremist activity.

The documents point to at least two types of risk. Extremists might use AI to plan attacks, recruit supporters, or create harmful material. They might also attack the data centers and other infrastructure that AI systems depend on.

More background

Some of these concerns existed before generative AI became popular. Reproductive health care centers and technology infrastructure, such as digital information systems, have long been targets for different kinds of extremist activity.

The Guardian reported that a 2021 warning from the Department of Homeland Security, the Federal Bureau of Investigation, and the National Counterterrorism Center discussed a plan by Seth Pendley, a participant in the January 6 riot and a member of a far-right militia, to attack an Amazon Web Services site.

"Pendley believed a successful attack on the Amazon facility would disrupt 70 percent of Internet service, expose perceived abuses of power, and lead to revolution," the warning said. Pendley pleaded guilty and was sentenced to 10 years in prison in 2021.

After ChatGPT launched in 2022, government agencies began examining how newer AI tools might make existing dangers worse.

The documents raised concerns about deepfakes, false information about elections, child sexual abuse material, and harmful propaganda. A 2025 memo from the Joint Regional Intelligence Center said the National Center for Missing and Exploited Children recorded more than 7,000 reports of child exploitation linked to generative AI between 2023 and 2025.

Agencies have also watched opposition to data centers. However, people who oppose these facilities because they use large amounts of energy, water, and land have warned that this monitoring could lead to intrusive police surveillance of people peacefully protesting.

What's being done?

In a public safety report, the FBI urged agencies to build relationships with AI companies and set up clearer ways to report dangerous activity. The Department of Homeland Security also said it regularly shares information about changing extremist threats with state and local partners.

Police departments may also be changing how they work internally.

A 2025 New York Police Department (NYPD) memo cited by the Guardian said agencies need to keep up with AI advances so they can respond to "increasingly sophisticated cyberattacks, other AI-enabled crime, and advanced malign influence campaigns."

Other law enforcement memos reviewed by the Guardian warned that AI-generated fake images and videos could target police and lead officers to respond with weapons when there is no real threat.

Law enforcement agencies want stronger safety protections and better ways to detect threats. At the same time, groups that defend civil liberties warn against treating general public criticism of AI or data centers as extremism. Better AI safeguards could help protect clinics, children, and communities. But surveillance that is too broad could also affect ordinary people who speak at public meetings.

Where can I learn more?

These related reports explain how AI can cause real-world harm and how officials are responding:

• Investigators found that AI chatbots helped teen users plan violent acts, raising further concerns about weak safety protections.

• Families said AI-related delusions drew loved ones deeper into paranoia and dangerous behavior.

• At an international summit, political leaders and technology executives warned that AI could make social problems worse.