TSA is looking at using AI to improve security screening processes

The Transportation Security Administration is looking at using artificial intelligence and other emerging technologies to bolster its security directive, with the agency’s top technology official saying these tools can improve the airport experience for both travelers and officers. 

In an interview with Nextgov/FCW, TSA’s Chief Technology, Data and AI Officer Matt Gilkeson said the agency “has already got the testing mechanisms in place and the stakeholder integrated project teams in place to be able to sustain and support continued exploration in machine learning and artificial intelligence tools.”

While many of TSA’s uses of AI are still in their infancy, Gilkeson said the agency’s deployment of automated systems, new algorithms and machine learning have demonstrated the effectiveness of using advanced technologies for security purposes. 

TSA and other agencies are increasingly looking at how AI can enhance their services, particularly after President Joe Biden released an executive order in October 2023 that outlined the government’s responsible and safe use of the emerging tools. The Office of Management and Budget subsequently released guidance in March establishing further safeguards around agencies’ uses of AI. 

The Biden administration’s policies have outlined an approach for agencies to securely incorporate these emerging technologies into their missions, and TSA has used that framework to begin formulating plans for the broader deployment of AI capabilities. 

Enhancing luggage x-ray screenings 

TSA’s current scanning technology identifies explosives when travelers’ luggage and personal items are run through x-ray machines, but officers have to be on the lookout for other prohibited items. The agency sees this problem as an area primed for change.

During an April House hearing, TSA Administrator David Pekoske said the agency is exploring how it can use AI to better pinpoint these barred objects. He said AI would be trained to detect all of TSA’s prohibited items, which would provide “a machine assist” to officers.

Gilkeson said the agency has a long list of prohibited items, and automated detection of these objects would “really aid the officers in the tedious work of reviewing images by starting to let the machines detect these items.”

“That's one of the biggest things for us, being able to expedite passengers' bags through the system by allowing the machines to do some more of that detection work,” Gilkeson added.

Both Gilkeson and Pekoske said that deployment of the technology would come after rigorous testing had been conducted, including with the Department of Homeland Security’s Science and Technology Directorate. TSA is a component agency of DHS. 

Pekoske told lawmakers last month that TSA is currently developing the system with an outside vendor and DHS S&T “to make sure that we're doing it in a proper way.” 

Departmentwide guidance helping AI initiatives grow

Building on the Biden administration’s directive regarding agencies’ uses of AI, DHS released an AI roadmap in March to help shape its use of the emerging capabilities. 

Gilkeson previously told Nextgov/FCW in April that DHS openness to AI is helping enable its components “to adopt the technology, explore it responsibly and look at what those use cases are.”

As part of DHS’s exploration of AI, Gilkeson said the agency is establishing AI testbeds that would allow TSA and other agencies under the department to “have some formal sandbox environments where we can continue the testing of these algorithms and these technologies.”

The agency’s willingness to test and explore these emerging capabilities has also extended to generative AI, with the department releasing an October 2023 memo that encouraged personnel “to responsibly use commercial products to harness the benefits of Gen AI and ensure we continuously adapt to the future of work.”

Gilkeson noted that DHS has “conditionally approved” its employees to use specific Gen AI tools “for publicly available information.”

“There's generative AI training that is required before you start doing this and supervisory approvals that are required, but the important component in that training is understanding how to critically analyze the information,” Gilkeson said. 

He added that both personnel and the general public should treat the output from these tools as “50% to 60% accurate,” and then use existing factual sources to verify the accuracy of the information. 

Gilkeson said that TSA is “exploring opportunities for how to create efficiency” with Gen AI tools, however, including looking at early use cases that align with DHS’s guidance and resources on using the tools. 

One of these early proposals, he said, includes “looking at how we take the tsa.gov publicly available information and create a responsive interactive chat engine for understanding what TSA’s policies are” that officers can use when it comes to identifying prohibited items or carrying out job tasks in certain circumstances. 

“We'll have a lot of testing that's got to be done to make sure that we understand how the systems perform,” Gilkeson said. “But, you know, this technology presents an incredible opportunity to do those things.”

Deployment of new capabilities

While TSA continues to explore the adoption of more advanced technologies, emerging tools — including those using machine learning and advanced algorithms — have already been rolled out in airports across the country. 

The adoption of a new algorithm based on machine learning principles, Gilkeson noted, has “improved security detection and decreased the pat-down rate” at security checkpoints. This capability was adopted roughly 18 months ago and has diminished pat-down searches by approximately 50%. 

In addition to making the screening process less physically intrusive, Gilkeson said the new system also included the adoption of a single scan button. The previous system had a male or female button, which required officers to make that determination. 

“By putting a scan button in place, we're also expediting and kind of moving the passenger experience forward in a positive way, because you're not having to make a gender call at all,” he said. “You're just scanning the passenger and resolving the alarms.”

TSA is also continuing to roll out facial biometrics at airports across the country to verify the identities of travelers. This technology, which has been deployed at more than 80 airports, compares real-time photos of individuals against their government-issued identifications. TSA is planning to roll this technology out to more than 400 airports in the coming years. 

TSA and DHS S&T officials previously told Nextgov/FCW about the security and privacy steps they are taking with the technology, including automatically deleting travelers’ photos after a scan has been completed and conducting testing on the equipment to ensure that it is working appropriately. They also noted that individuals can freely opt-out of the facial scanning process without consequences, and that signage stating that information is displayed at airports with the technology. 

Some lawmakers and privacy experts, however, have voiced concerns about the adoption of this next-generation technology and the possible impact it will have on Americans’ privacy moving forward, including the possible slippery slope it could create by normalizing surveillance tools. 

A bipartisan group of senators have proposed an amendment to the Federal Aviation Administration reauthorization bill that would freeze TSA’s deployment of facial biometrics at additional airports while Congress reviews the program. The FAA’s current authorization expires on May 10.