The most terrifying part of artificial intelligence is an omnipresence that can reach anyone even without them knowing they were targeted, according to one of the FBI’s top tech specialists.

Kathleen Noyes, who leads the FBI’s next-generation technology and lawful access section, said Tuesday she categorizes threats as most dangerous and most likely — and she labels AI as the most likely threat.

“Every single one of you has … brushed it,” Ms. Noyes said at an AI conference hosted by General Dynamics Information Technology. “You may not even know that you brushed it. You may not even know that perhaps you were the target. So I think that, to me, is the scariest part.”

---

---

Ms. Noyes thanked many of the AI event’s attendees in Washington, who she said were responsible for cutting-edge tech deployments.

She urged people to be very careful about using AI and she said the bureau has watched people put proprietary information into generative AI tools, which risks exposing sensitive data.

Ms. Noyes has previously tussled with privacy and civil liberties advocates, who expressed concerns about the federal government’s hands-on approach to tech, particularly over law enforcement officials’ desire to create digital backdoors weakening encryption.

But FBI agents say the exploitation of AI by criminals is making their job much harder.

The bureau’s Cynthia Kaiser said digital adversaries, ranging from China, Iran, North Korea, and Russia to cybercriminals, use AI to be more efficient, to make their lies more believable, and to avoid detection.

“They’re developing code to evade malware detection to learn about a system so they can just stay hidden more,” Ms. Kaiser said at the conference. “And really, I think what that comes down to is you have a crop of adversaries who are becoming at least mildly better. A beginner hacker can go to the intermediate level, and even the most sophisticated adversaries can be more efficient.”

While citing the dangers, the FBI is not above using AI to improve its own operations.

Ms. Kaiser, deputy assistant director of FBI’s cyber division, said the bureau is deploying AI to improve its processes for digging through tips.

When people reach out to the FBI with tips, humans are no longer the only ones reviewing the message to sound the alarm for potential criminal activity.

“There’s always going to be a person taking that down, but people miss things,” Ms. Kaiser said. “After it’s looked at by a human, after it’s flagged one way or another, we’re using natural language processing models to also go over the synopsis of the text of what that phone call or online [tip detailed] to see [if we missed] something.”

Ms. Kaiser said the AI tool is trained on the FBI’s expertise at sorting through incoming tips and she stressed that there would always be a human in the loop for FBI’s handling of its tip line.