FBI leaders say more workers need to be trained in AI

As the latest developments in artificial intelligence grow more readily available, leaders at government agencies like the Federal Bureau of Investigation are trying to prevent hackers from succeeding with these new tools. 

But this is no easy feat. Accessible technology, like AI, is especially dangerous and more complex than other tech seen before. Government and technology leaders attested to this reality at an early June event that consulting company General Dynamics Information Technology (GDIT) hosted in DC. 

Anyone with a computer can exploit AI, and it’s making adversaries become much more efficient, they explained. 

“You pretty much can go into a generative AI tool and say, ‘Hey, help me not write malicious code. Do the reverse,’” said Kathleen Noyes, the FBI’s section chief of next-generation technology and lawful access. “And suddenly: ‘Here are all the things you would want to avoid if you wanted to write malicious code.’”

In addition to malicious code, AI can be used to create deep fakes, develop phishing scams and spread false information, panelists explained. The international NGO World Economic Forum backs the speakers’ concern by classifying AI as a major emerging global risk, particularly through its use to distort information. 

To stay ahead of bad actors, both domestically and abroad, leaders emphasized that workers need to be well-versed in AI. When developing these innovations, clarity is necessary, too. 

AI needs to be key to workforce training

There’s a pressing demand for workers with AI skills across the public and private sectors. Because of that, making an effort to “upskill” workers is necessary, said FBI section chief Noyes.

“We have to invest in our workforce,” Noyes said at the event. “We have to get everyone a baseline of knowledge.” 

FBI leadership is currently experimenting with a “Shark Tank”-type program to foster innovation within the agency. Under this program, named after the famous ABC show, employees have 90 days to develop and prove a concept. At the end of that period, the agency will evaluate cost, skills needed and how it would work if integrated into processes, explained the FBI’s interim CTO David Miller. 

“It becomes this huge educational opportunity,” Miller said. “It allows us to have really strategic innovation in doing outcomes.”

The FBI has also been rolling out different AI use cases, including using natural language processing models to help maintain the agency’s tip line. People still answer the phone while AI helps review the calls to see if anything was missed. 

The average person may miss or overlook key information, and the models are an example of AI “making us so much safer,” said Cynthia Kaiser, the deputy assistant director of the FBI’s cyber division.  

A transparent approach is essential

Justin Williams, the deputy assistant director for the agency’s information management division, said that when developing these new AI tools, there needs to be an explanation as to why it’s being used and how it was created. 

There’s always the chance that what’s being used at the agency, AI or otherwise, will have to be defended in court, Williams said. An innovation may also feature in the news, which requires a clear public explanation. 

Williams said he’s used several different generative AI tools on the job, but when asking platforms similar questions, he’s received slightly different responses. This lack of stability needs to be noted, Williams said. 

The FBI developed an AI ethics council in 2021 to evaluate AI use cases, Noyes explained, which has been a key part of the agency’s plan to remain transparent. The FBI evaluates risks and prioritizes use cases where there’s a clear answer that AI could be useful. 

“It can’t be a black box,” Noyes said. “We need some transparency and accountability for knowing when we’re invoking an AI capability.”