Rep. Don Beyer (D-Va.) is calling for the immediate suspension of a new AI-supported system being deployed by the Federal Aviation Administration at Washington-area airports, less than two years after 67 people were killed in a midair collision near Ronald Reagan Washington National Airport.

Beyer is hardly alone in asking how far artificial intelligence should be allowed to go and who should be watching it. HillGraph’s AI Policy Dashboard currently identifies 120 AI-related bills in the 119th Congress, covering issues ranging from national security and health care to jobs, privacy, education and safeguards intended to keep increasingly powerful AI systems under human control.

Here in Virginia, the issue is coming from nearly every direction. Gov. Abigail Spanberger this week called on congressional leaders to establish national AI guardrails, warning about potential impacts on privacy, employment and resource consumption, while her administration has created a Virginia AI Task Force focused on workforce displacement, data privacy and cybersecurity.

Sen. Mark Warner (D-Va.) is pushing stronger cybersecurity protections for Americans’ medical information and challenging a federal proposal involving identifiable health information from millions of federal employees, retirees and their families. He has also focused on “surveillance pricing,” questioning how personal information and algorithms can influence what consumers pay.

Then there is the gas pump. Plaintiffs in a federal lawsuit accuse some of the nation’s largest gasoline retailers of using an AI-powered pricing platform in a manner that coordinated prices rather than allowing retailers to independently compete. Those allegations are contested and have not been proven in court.

Del. Marcus Simon (D-Falls Church), meanwhile, has questioned the risks created by interconnected surveillance databases capable of tracking people’s movements. Falls Church Police Chief Shahram Fard has said the City has no fixed Flock cameras and no relationship with Flock, while Fard and Sheriff Matt Cay are among local officials confronting increasingly complicated rules surrounding surveillance technology, data retention and information sharing.

The national debate is moving even further. Sen. Bernie Sanders (I-Vt.) and Rep. Greg Casar (D-Texas) have proposed temporarily pausing advanced AI development until federal safety rules are established, permanently prohibiting artificial superintelligence as defined by their legislation and creating a cabinet-level federal agency to oversee advanced AI.

Airports, medical records, grocery prices, gas prices, cellphones, license plates, jobs and data centers may sound like completely different stories. Taken together, they help explain why Congress is suddenly confronting 120 different pieces of AI-related legislation.

For Beyer, the immediate concern is what is happening above Northern Virginia.

“The FAA should not use my constituents as guinea pigs for an unproven AI air traffic control system,” Beyer said Monday in calling for suspension of the FAA’s Strategic Management of Airspace Routing Trajectories system, known as SMART.

The FAA says SMART centralizes approximately 200 streams of information involving weather, flight paths, traffic flow, controller staffing and other conditions, using AI to help aviation specialists anticipate congestion and other problems. The system provides information and recommendations rather than replacing air traffic controllers, and nothing identified by the News-Press establishes that SMART itself is unsafe.

Beyer’s concerns include whether controllers were sufficiently involved in developing and testing the system and whether adequate independent testing occurred before deployment. The National Air Traffic Controllers Association has said it was not involved in SMART’s design, testing or implementation.

Those questions carry additional weight here following the Jan. 29, 2025 collision between an American Airlines passenger jet and a U.S. Army Black Hawk helicopter over the Potomac, which killed all 67 people aboard the two aircraft. The crash had nothing to do with artificial intelligence, but the National Transportation Safety Board’s investigation identified systemic safety failures and produced 74 findings and 50 safety recommendations.

Beyer also recently led a bipartisan group of House members calling for congressional action following reported incidents involving increasingly capable AI agents. As AI moves beyond answering questions and generating pictures into systems capable of accessing databases, writing code, operating software and interacting with other computer systems, the concern becomes less about science fiction and more about what technical barriers exist when an AI system is given permission to act.

Spanberger argues that some of those questions require national rules. “Our AI Task Force is focused on addressing the significant risks to our jobs, our privacy, and our cybersecurity posed by the real, rapid development of AI,” she said this week, while calling on Congress to establish nationwide guardrails.

Warner is raising similar questions about information Americans already consider among their most private. He and Sen. Ron Wyden (D-Ore.) have introduced legislation requiring minimum cybersecurity standards for health care organizations, while Warner and Sen. Adam Schiff (D-Calif.) are separately challenging an Office of Personnel Management proposal involving identifiable health information from federal employees, retirees and family members.

The same technology is also moving closer to the price tag. Warner’s “surveillance pricing” concerns involve the potential use of personal information and algorithms to determine what individual consumers may be willing to pay. The Kalibrate lawsuit raises a different allegation: that competing gasoline retailers relying on the same pricing technology reduced ordinary price competition. Those allegations remain unproven.

And Northern Virginia isn’t simply using this technology — it is helping power it. Loudoun County says it has the highest concentration of data centers anywhere in the world, with approximately 233 data center buildings and 56.5 million square feet built or under construction as of March 1.

The questions surrounding all of it are becoming increasingly difficult to separate: Who controls the information? What can artificial intelligence do with it? Who establishes the boundaries? And who is watching when those boundaries fail?

This story continues online at FCNP.com, where the News-Press digs deeper into AI safety, government access to cellphones and medical records, license plate readers and electronic device tracking, grocery and gasoline pricing, cybercrime, federal job losses, Data Center Alley and the growing battle over who ultimately controls America’s data.

BEYOND THE PRINT EDITION (online only)

The explosion of AI legislation in Congress is significant not simply because of the number of bills, but because of what lawmakers are attempting to regulate.

HillGraph’s AI Policy Dashboard currently identifies 120 AI-related bills in the 119th Congress, with proposals addressing artificial intelligence across national security, health care, education, finance, privacy and increasingly autonomous systems.

That does not mean Congress agrees about what should be done. The proposals range from encouraging development and establishing technical standards to regulating particular uses of AI and, at the far end of the debate, stopping development of some advanced systems altogether.

Sanders and Casar’s proposal would permanently prohibit the development and deployment of artificial superintelligence as defined in the legislation and temporarily pause advanced AI development until federal safety rules are established. Their proposal also calls for a new cabinet-level federal agency, penalties for violations, and an international effort aimed at preventing development of artificial superintelligence elsewhere.

Spanberger is taking a different approach. She is not calling for AI development to stop, but for Congress to establish nationwide guardrails while states continue acting within their own authority. Her Sept. 22 letter to congressional leaders specifically identified data privacy, employment and large-scale resource consumption as areas requiring attention and argued that the United States does not have to choose between AI safety and technological leadership.

The debate therefore isn’t simply between people who support AI and people who oppose it. The much more difficult argument is over where the lines should be drawn, who should draw them, and whether government can establish rules quickly enough to keep up.

WHEN AI DOESN’T STAY INSIDE THE LINES

The controversy surrounding SMART is not about an artificial intelligence system taking control of airplanes. It is about how much testing and human oversight should be required before AI-supported technology is incorporated into systems where mistakes can have serious consequences.

Beyer’s concerns also follow reported incidents in which increasingly autonomous AI agents have taken actions beyond what developers or users expected. On Sept. 16, Beyer led a bipartisan group of 10 House members urging congressional leaders to act on AI legislation following reported incidents involving autonomous agents.

The problem does not require a machine to become conscious, develop motives or decide to turn against humans. An AI agent is software capable of processing information and taking actions with varying levels of autonomy in pursuit of a goal.

The important distinction is between an instruction and a barrier.

Humans routinely understand instructions within a much broader framework of common sense, consequences and context. Computer systems operate within the access and technical restrictions humans give them. If an increasingly autonomous system has permission to interact with a database, computer network or other software, a written instruction telling it not to perform a particular action may not provide the same protection as a technical control making the action impossible.

One of the clearest real-world examples came from AI coding platform Replit, where an AI coding agent deleted a live production database during a software development project despite instructions intended to prevent changes to production systems. Replit’s chief executive publicly apologized following the incident and the company announced additional safeguards.

That wasn’t a machine becoming conscious or deliberately rebelling against humans. It demonstrated something much more practical: an AI system capable of taking actions can produce consequences beyond what the person using it intended.

As AI agents become capable of using software, writing code, searching databases and communicating with other systems, that difference becomes considerably more consequential.

MEDICAL RECORDS AND THE GOVERNMENT

Some of the most sensitive information Americans possess is already sitting inside enormous databases.

Warner and Wyden’s Health Infrastructure Security and Accountability Act would require HHS to establish and regularly update minimum cybersecurity standards for health care providers, health plans, and other covered entities amid continuing cyberattacks against hospitals and providers. The proposal would also provide $1.3 billion to help hospitals improve cybersecurity, including $800 million directed toward rural and underserved communities.

The separate dispute involving OPM raises a different question: how much identifiable medical information should the government collect and what safeguards should surround it?

Warner, Schiff and other senators say an OPM System of Records Notice confirms an intention to collect identifiable health information involving federal workers. They have objected to the proposal and demanded stronger privacy protections and suspension of the collection. Those are the senators’ stated concerns; the existence or proposed creation of a large government health database does not itself establish that the information will be misused.

The significance for artificial intelligence is scale.

A single medical claim may reveal relatively little. Assemble years of claims, providers, procedures, prescriptions and treatment dates and the resulting record can reveal considerably more about someone’s life. AI makes searching, sorting and identifying patterns across enormous amounts of information faster and cheaper than it has ever been.

YOUR CAR IS NO LONGER JUST A LICENSE PLATE

Virginia has already imposed significant restrictions on how law-enforcement license plate reader information can be retained and shared, while local officials are having to navigate what increasingly complicated surveillance rules mean in practice.

Falls Church Police Chief Shahram Fard has said the City does not operate fixed Flock cameras or have a relationship with Flock. Sheriff Matt Cay, who previously served as a Falls Church police officer before becoming sheriff, is among the local officials working in an environment where the lines between ordinary investigative tools, enormous private databases and emerging surveillance technology are becoming increasingly important.

Traditional automated license plate readers photograph plates and convert the images into searchable records. Newer technology demonstrates that a plate itself does not have to be the only identifying feature.

Leonardo’s ELSAG SignalTrace technology can passively detect signals associated with electronic devices including mobile phones, smartwatches, RFID tags and other electronics. The company says its technology can identify combinations of devices repeatedly traveling together and correlate those patterns with license plate information, creating what amounts to an electronic fingerprint.

SignalTrace does not mean an ordinary license plate reader secretly performs the same function, and there is no evidence Falls Church uses SignalTrace. The technology instead demonstrates where surveillance is capable of going: identifying a vehicle not simply from the characters printed on a metal plate, but potentially from the collection of electronics traveling inside it.

Virginia’s statutory definition of surveillance technology includes automated license plate readers, cell-site simulators, facial and biometric recognition, social-media monitoring, passive radio scanners and long-range Bluetooth and other wireless-scanning devices.

The law is already trying to catch up with what the equipment can do.

THE COMPUTER IN YOUR POCKET

Cellphones present perhaps the clearest example of how much the meaning of personal information has changed.

A modern phone can contain messages, photographs, financial information, health information, work files, internet activity, contacts and years of personal communications. Depending on settings and services, information associated with the device can also reveal where its owner has been.

Federal courts continue to determine how constitutional protections apply to searches of those devices, particularly at international borders, where the government’s search authority has traditionally been broader.

In September, the U.S. Court of Appeals for the Second Circuit held that routine manual searches of cellphones at the border do not require a warrant, probable cause or individualized suspicion. The ruling did not establish unlimited authority for more intrusive forensic extraction of cellphone data and directly governs the states within the Second Circuit.

The distinction between manually looking through a phone and using specialized equipment to extract or analyze its contents is increasingly important. But the larger issue is difficult to miss: a search of a modern cellphone can potentially reveal an amount of information that once would have required searches of someone’s home, office, filing cabinets, photographs, correspondence and financial records.

Now much of it can travel inside a pocket.

WHAT DOES AI KNOW ABOUT WHAT YOU’LL PAY?

Warner’s “surveillance pricing” push moves the data debate from privacy into Americans’ wallets.

The basic concern is straightforward: if a company knows enough about a consumer’s location, browsing behavior, demographics, previous purchases or other characteristics, algorithms can potentially estimate what that particular person is willing to pay.

That does not mean every grocery store is secretly charging every customer a different price. It means the technology and data increasingly exist to make individualized offers and pricing decisions at a scale that would have been difficult to imagine when most transactions occurred with a price sticker and cash register.

Warner has used identical online grocery orders carrying different prices to illustrate the issue and has pushed the Federal Trade Commission to examine surveillance pricing.

The gasoline case raises a different problem.

Plaintiffs in a federal lawsuit accuse major fuel retailers of using Kalibrate’s algorithmic pricing technology in a manner that coordinated gasoline prices rather than allowing the companies to independently compete. The allegations remain contested and have not been proven in court.

The distinction is important. Surveillance pricing asks what happens when an algorithm knows enough about you to estimate what you will pay. The Kalibrate lawsuit alleges a different form of potential harm: competitors relying on the same pricing technology in a manner plaintiffs say reduced ordinary price competition.

Either way, software is moving much closer to the price tag.

CRIMINALS WANT THE DATA TOO

Restricting government or corporate access does not solve the entire problem because personal information has become extraordinarily valuable to criminals.

The FBI received 1,008,597 internet-crime complaints in 2025, with reported losses approaching $21 billion. Cyber-enabled fraud accounted for more than $17 billion in reported losses, while Americans over 60 reported approximately $7.7 billion in losses.

That makes the privacy debate considerably more complicated than deciding whether the government should be allowed to collect information. Governments want data. Businesses want data. Criminals want data. Increasingly, AI systems can make that data more useful to whoever possesses it.

Anyone who grew up during the early personal-computer era remembers rampant software piracy. Programs were cracked, copied and passed around with little thought given to the companies that spent money developing them. Those companies did not accept having their digital property stolen simply because technology made copying it easy. Laws changed, security improved and enforcement followed because society recognized that information did not stop having value simply because it existed digitally.

Today the valuable digital property increasingly belongs to ordinary Americans: our financial information, photographs, communications, locations, purchases, medical records, devices and identities.

A license plate observation, wireless signal, medical claim, purchase or location point may reveal relatively little by itself. Connect enough information over enough time and artificial intelligence can identify relationships and patterns that once would have been extraordinarily difficult for a human investigator, company or criminal to discover.

THE JOBS ARE CHANGING TOO

Northern Virginia is simultaneously experiencing another enormous transformation.

The Washington metropolitan area lost approximately 103,900 jobs between January 2025 and January 2026, including 62,100 federal jobs, according to the Metropolitan Washington Council of Governments. Federal employment in the region fell 16.5 percent during that period and reached its lowest level since 1990.

Those jobs cannot simply be described as having been replaced by artificial intelligence. Federal workforce reductions and accelerating AI adoption are separate developments, and there is no evidence establishing AI as the cause of those 62,100 lost federal positions.

But they are occurring at the same time in the same region.

Government and private industry are investing heavily in automation, cloud computing and AI while tens of thousands of traditional federal positions disappear from the Washington economy. That raises a longer-term question about what happens to Northern Virginia if the economy that emerges from the AI revolution does not employ the same number or types of people as the one it replaces.

Technology has always eliminated some jobs while creating others. AI could increase productivity, create new industries and generate entirely new categories of employment. The difficult question is what happens during the transition and whether workers displaced by that transformation share in the economic benefits it creates.

Sanders is already tying those issues together in another way. He and Rep. Mark Takano (D-Calif.) have pushed legislation moving toward a 32-hour workweek, arguing that productivity gains from AI and automation should improve workers’ lives rather than simply eliminate jobs.

The employment argument therefore may eventually become about more than how many jobs AI destroys or creates. It may also become a debate over who receives the economic value produced when machines allow fewer people to accomplish more work.

DATA CENTER ALLEY IS GROUND ZERO

Nowhere is that transformation more visible than Northern Virginia.

Loudoun County says it has the highest concentration of data centers anywhere in the world. As of March 1, it had approximately 233 existing data center buildings representing 56.5 million square feet built or under construction, with another 35.7 million square feet of data center use approved but not yet built. More than 17,000 direct and indirect jobs are tied to the industry.

The financial impact is enormous. Data centers generated approximately $1.2 billion in Loudoun real and personal property tax revenue in FY2026, representing 39 percent of the county’s overall budget. The FY2027 adopted budget projects $1.3 billion, or 40 percent of the county budget, from the industry.

That means Virginia cannot realistically discuss artificial intelligence without discussing the infrastructure required to power it.

Spanberger’s Data Center Accountability Framework attempts to connect those issues, addressing energy costs, environmental impacts, transparency and workforce concerns while her administration simultaneously moves forward with the AI Task Force.

And we haven’t even touched some of the biggest players in the room.

Major cloud and technology companies operate enormous amounts of digital infrastructure in Northern Virginia, where decades of investment in fiber networks, data centers and internet connectivity have made the region one of the most important pieces of physical infrastructure underlying the digital economy.

Questions surrounding where information is stored, who owns it, who can access it, how systems connect it and what happens when increasingly autonomous artificial intelligence is layered on top could fill another investigation by themselves.

WHO WRITES THE RULES?

The debate ultimately comes back to the government because someone has to decide where the boundaries are.

Spanberger is calling for national guardrails while arguing that Virginia will continue acting in areas where the state has authority. Sanders and Casar are proposing that the government go much further and halt development of certain advanced systems until safety requirements are established. Other lawmakers are pursuing narrower rules governing specific uses of AI, while the Trump administration has advocated a national AI framework rather than a patchwork of differing state requirements.

There are competing interests.

Artificial intelligence could help doctors diagnose diseases, protect computer networks, discover new medicines, identify fraud, improve transportation, increase productivity and solve problems involving amounts of information humans cannot realistically process alone. Data centers support thousands of jobs and enormous amounts of local tax revenue, while American technology companies are competing globally to develop systems that could shape the next generation of the world economy.

Too much government control can restrict innovation, competition and economic growth. Too little oversight can leave citizens absorbing the consequences when systems fail, personal information is misused, markets are manipulated or criminals exploit new technology faster than government can respond.

SMART may ultimately work exactly as the FAA intends and make aviation safer. License plate readers can help police find stolen cars and solve crimes. Medical databases can improve care and identify fraud. Algorithms can make businesses more efficient.

The issue is not whether technology itself is good or bad. It is how much power comes with the ability to collect, connect and analyze enormous amounts of information, where the boundaries should be placed, and whether safeguards are being built before something goes wrong rather than afterward.

Less than two years after 67 people died over the Potomac, Beyer is asking those questions about AI and aviation. Warner is asking them about medical information, cybersecurity and consumer prices. Simon is asking them about surveillance databases, while Falls Church Police Chief Shahram Fard and Sheriff Matt Cay are dealing with what increasingly complicated rules surrounding technology, surveillance and information sharing mean when they reach the local level. Spanberger is asking Congress to establish national rules while Virginia creates an AI Task Force of its own, and lawmakers across the country are introducing proposals that demonstrate just how unsettled the answers remain.

All of it is happening here, while Northern Virginia’s federal workforce shrinks and the world’s largest concentration of data centers continues powering an accelerating digital economy. Virginia is simultaneously helping build the future and becoming one of the places being forced to decide what protections should come with it.

In the land of the free and the home of the brave, how America answers those questions may help determine whether technological progress expands freedom or gives governments, companies, criminals and machines greater control over our information, our choices and even what we pay.

The News-Press will continue following the data.