Kash Patel says that AI use at the FBI has 'increased by 605%' since he became director — claims that every major tech player is 'embedded' in the agency

Kash Patel Says AI Use at the FBI Has 'Increased by 605%' Since He Became Director

FBI Director Kash Patel has claimed the bureau's use of artificial intelligence has risen 605 percent under his leadership, and that every major tech company in the world is embedded at the agency.

The remarks came Saturday on Fox News' "Saturday in America" with host Kayleigh McEnany.

What He Said

"AI, when used lawfully, is a critical tool to triage data," Patel said. "For far too long the institutional morass in D.C. rejected the advancements of technology. I have increased the use of AI in the FBI by 605%."

He said the technology had helped prevent school shootings in North Carolina and about half a dozen other states since he took office.

On what comes next: "What we're about to do is work with our private sector partners and work with our state and local partners and look at databases and collect information and get DNA and then get these threats back out to the field and stop people from killing children at schools."

Patel described the effort as apolitical, noted that Senate Democrats had not raised AI during his Judiciary Committee hearing on Tuesday, and said the bureau had made the most technological advancements in its history over the past 18 months.

The 605 percent figure also appeared in his prepared remarks for congressional oversight testimony this week, where he attributed it to expanded use of AI for threat detection, triage, analysis and processing.

The Number Has No Denominator

The central problem with the claim is that 605 percent of what is not stated anywhere.

There is little public data on how much AI the FBI used before Patel took office, or in what ways. Without a baseline, a sevenfold increase could describe anything from a substantial institutional transformation to a modest expansion from a very small starting point.

It is also unclear what is being measured. Number of queries? Licences purchased? Cases where an AI tool was involved at any stage? Compute spend? Each of those would produce a different figure, and a percentage this precise implies a specific metric that has not been disclosed.

The specificity is itself worth noting. Round numbers get rounded; 605 percent reads as though it came from a calculation. What calculation has not been made public.

What the FBI Actually Appears to Be Doing

Setting the percentage aside, there is evidence of genuine deployment.

Patel has described the bureau using new AI tools to generate call transcriptions, produce concise synopses, and help correlate contacts with other received complaints. He noted that messages collected under a search warrant can take weeks for a team of analysts to process — work AI handles quickly.

That is a credible and fairly mundane application. Transcription, summarisation and entity correlation across large document sets are among the most reliable things current models do, and evidence processing is exactly the kind of volume problem they suit.

It is also considerably less dramatic than the framing around school shootings suggests.

The Attribution Problem

Patel told Fox News' Sean Hannity in May that "we stopped a school massacre in North Carolina because we got a tip and we were able to triage it with artificial intelligence."

Read carefully, the AI's role in that account is triage — sorting an incoming tip for priority. The tip came from a human. The investigation and intervention were conducted by agents.

Crediting a prevented attack to AI in that chain is a stretch. Triage software that moves an urgent report up a queue is useful, and it may well have saved time that mattered. It is not the same as the technology identifying the threat.

There is also no independent way to verify the count. Prevented attacks are inherently unverifiable from outside — the counterfactual cannot be observed, and the underlying cases are not public.

The 'Embedded' Claim

Patel has separately said he has "every major tech company in the world embedded in the FBI, rebuilding our internet capabilities."

That phrasing raises questions the interview did not address. Embedded in what capacity — as contractors, as secondees, as vendors with on-site support staff? Under what contracts, with what oversight, and with what access to bureau data?

Private-sector technical staff working inside a federal law enforcement agency is not unusual in itself. Large systems integration work routinely involves contractor personnel. But the combination Patel describes next — working with private sector and state and local partners to look at databases, collect information and gather DNA — is a description of expanded data aggregation across jurisdictions.

That is the part deserving scrutiny, and it is the part that got the least in a Fox News segment about school shootings.

The Timing

The claim lands in a week where the AI conversation in Washington has been dominated by calls for restraint.

Anthropic CEO Dario Amodei warned this month that the pace of AI development must be slowed. Elon Musk and Sam Altman backed the call. Mark Zuckerberg publicly rejected it. Congress remains divided, and no federal framework exists.

An agency head touting a sevenfold increase in AI deployment for threat detection sits somewhat awkwardly alongside that debate — not because law enforcement should not use the technology, but because the oversight framework governing how it uses it has not been written.

Sources: Tom's Hardware, Mediaite, Benzinga, Irish Star