Van Hollen Joins McClain Delaney, Colleagues in Call to Pause FAA’s AI Tool at Washington-Area Airports

U.S. Senator Chris Van Hollen (D-Md.) joined Congresswoman April McClain Delaney (D-Md.), along with Representatives Alexandria Ocasio-Cortez (D-N.Y.), Don Beyer (D-Va.), Kweisi Mfume (D-Md.), Suhas Subramanyam (D-Va.), and Sarah Elfreth (D-Md.), in urging the Federal Aviation Administration (FAA) to pause use of its new artificial intelligence-enabled air traffic management tool until the agency answers questions about safety testing, implementation, and oversight.

The letter to Transportation Secretary Sean Duffy and FAA Administrator Bryan Bedford raises concerns about the September 21 deployment of the Strategic Management of Airspace, Routes, and Trajectories (SMART) tool at Ronald Reagan Washington National Airport, Washington Dulles International Airport, and Baltimore/Washington International Thurgood Marshall Airport.

At a time when AI safety concerns are sky-high, the lawmakers warn that the FAA has not provided sufficient transparency around its decision-making process prior to deployment, nor the functioning of SMART's underlying AI systems. Congress, and the public, deserve to know how AI systems were tested and how they will influence decisions before being deployed in safety critical systems.

The letter also raises concerns about reports that frontline aviation safety workers received only days of notice, or no notice at all, before deployment. With air traffic controllers already facing staffing pressures, the lawmakers stress that new technology must be introduced with adequate training and clear procedures for human oversight.

The lawmakers request answers in three areas:

• Safety testing: What evaluations were completed before deployment, how the system performs during severe weather or outages, and whether any safety review requirements were waived during the contracting process.
• Consultation and training: What input the FAA sought from airport authorities, air safety workers, and airlines; what training personnel received; and how the tool should influence decision-making.
• Evaluation and accountability: How the FAA will measure performance during the initial 90-day evaluation, what standards must be met before nationwide expansion, and who is responsible if a recommendation contributes to a safety incident.

The letter calls for a pause until lawmakers can better understand the FAA's decision-making, rollout process, and evaluation plan. The co-leads emphasize that modernizing air traffic control must prioritize safety, transparency, and the expertise of the people responsible for keeping our skies safe.

The full text of the letter is availble here and below.

Dear Secretary Duffy and Administrator Bedford:

We write to express our concerns regarding the Federal Aviation Administration's (FAA) rollout of the AI-enabled Strategic Management of Airspace, Routes, and Trajectories (SMART) tool to help manage air traffic above the National Capital Region (NCR) at the Ronald Reagan Washington National (DCA), Baltimore/Washington International Thurgood Marshall (BWI), and Washington Dulles International (IAD) airports. The NCR is one of the busiest airspaces in the country and is still reeling from the impacts of the tragic midair collision of a passenger flight and military helicopter over DCA, in which 67 people died. We are incredibly concerned by the use of the NCR region to pilot AI tools in safety-critical aviation systems, especially as we have not seen any safety evaluation and testing reports, nor have we been made aware of any efforts to train aviation safety personnel on how best to use the tool. Given extensive recent reporting on AI cybersecurity incidents and other unpredictable behavior from AI systems, we are deeply concerned about the integration of any AI-enabled system into air traffic control systems, where decisions are literally a matter of life or death.

We understand that the SMART tool was built through a 12-year, $875 million contract. Yet here has been little transparency from either the Department of Transportation (DOT) or the FAA on how the AI systems underlying the SMART tool were tested, how it is being integrated into air traffic control (ATC) systems, and how the tool's recommendations may impact already overburdened airspace and air traffic controllers. Nor has information been shared with local officials as to why the NCR, the region in which the deadliest U.S. air disaster in nearly 25 years occurred just last year, was chosen as the pilot test region. This, and other recent tragic aviation accidents such as the fatal March runway crash at LaGuardia Airport in New York, have highlighted the need for more investment in safer skies. need to modernize ATC systems - Congress appropriated $12.5 billion last year to do just that. However, any upgrades and changes to ATC systems must be made with safety as the first priority. We recognize the urgent need to modernize ATC systems - Congress appropriated $12.5 billion last year to do just that. However, any upgrades and changes to ATC systems must be made with safety as the first priority.

However, we have heard from several sources that frontline aviation safety workers were notified of the SMART tool rollout mere days in advance or not at all, giving them no time to learn how the system works, how it is incorporated into their workflow, and what oversight and accountability mechanisms maintain aviation safety standards. There is already a shortage of air traffic controllers, and we are concerned that integrating this new technology into the workflows of already-overworked aviation safety personnel could have deleterious consequences. The air traffic control systems and processes currently in place are complex and we don't know what impact AI-enabled tools might have on these systems and resulting safety. Decades of research shows that people tend to over-rely on automated recommendations, a phenomenon known as automation bias, particularly under overworked conditions. Adding new, untested layers to these systems, especially those that are not yet "ready to immediately transform flight operations", according to airline industry officials, must be done in partnership with all stakeholders in the aviation safety ecosystem.

We understand that the rollout of the SMART AI tool in the NCR on Monday, September 21, is part one of a phased rollout that begins with a 90-day evaluation period before broader, nationwide deployment. However, given the egregious lack of transparency around the initial rollout, especially at a time where there is so much public concern around the use of AI in critical systems, we ask that you pause any use of the SMART tool until we can better understand your decision-making, rollout process, and evaluation plan. We ask that you respond to the following questions no later than October 19, 2026:

Pre-Deployment Testing & Evaluations

1. Did the FAA make a formal In-Service Decision before SMART began operational use?

a. If so, which official and office made it, on what date, and on what basis?

b. Was the decision made within the FAA's Air Traffic Organization, or did Department of Transportation leadership direct or approve the launch date?

2. Was a Safety Risk Management evaluation conducted under the Air Traffic Organization's Safety Management System, and was a Safety Risk Management Document completed and signed? If so, please provide it.

3. Were any standard acquisition or safety review steps waived or truncated to meet the September 21 launch?

4. What AI models does SMART use and how has the FAA validated their accuracy? Please provide detailed information regarding what pre-deployment testing and evaluations were conducted on the AI models underlying SMART, as well as measures of SMART tool

5. What factors did you consider in your decision to pilot this system in the National Capital Region rather than other regions? How was the final determination made and who made the determination?

6. OMB Memorandum M-25-21 categorizes AI used in "traffic control systems and other systems controlling physical transit" as high-impact, requiring pre-deployment testing, an AI impact assessment with independent review, and sufficient operator training. Has the Department determined SMART to be high-impact AI? If so, please provide documentation of the practices completed under Section 4(b). If the Department has determined it is not high-impact, or waived any practice, please provide the written determination required under Section 4(a).

Stakeholder Consultation & Training

7. What consultations did you conduct before deploying the SMART tool? What input did you solicit, and what feedback did you receive? Specifically include, at minimum, information regarding consultations with:

a. the Metropolitan Washington Airports Authority

b. the Maryland Aviation Administration,

c. the National Air Traffic Controllers Association, and

d. airline industry officials.

8. What training did aviation safety personnel receive prior to deployment of the SMART tool?

9. How will the tool be integrated into existing ATC systems and processes?

10. What weight are controllers and traffic management specialists expected to give SMART recommendations when making decisions as compared to other factors? Do they have the ability to manually override SMART recommendations, including after local facility leadership has accepted them? If so, what is the process, and how are those overrides documented?

11. How are recommendations, and whether they were accepted or overridden, tracked and evaluated? Are they recorded in the National Traffic Management Log or a similar system?

12. Did the FAA satisfy any applicable collective bargaining or labor consultation obligations before deploying SMART? Please describe what notice was provided to the exclusive representatives of affected employees and when.

Assessments & Oversight

13. What data will be tracked in the 90-day evaluation?

14. What specific metrics will determine whether the 90-day evaluation is a success? How will progress toward that goal be measured, and who will conduct the evaluation?

15. What is the timeline for further implementation of this system and what benchmarks will determine if the deployment process should continue?

16. What is your plan for ensuring public transparency and trust-building with the flying public and aviation industry as you deploy the SMART tool?

17. Who will provide oversight of the system?

18. What system do you have in place to enable independent oversight of the SMART tool, and who is responsible for conducting oversight (including but not limited to review by the National Transportation Safety Board following an incident)? Please provide details regarding your audit process of overseeing how the SMART tool reaches its recommendations, what actions are taken or not taken in response to these recommendations, and any safety implications of those decisions.

a. In the event that a SMART recommendation contributes to a loss of separation or other safety incident, who bears responsibility - the controller who accepted the recommendation, the FAA, or the developer? Has the FAA made any determination on this question?

Sincerely,