The Senate questions Flock (with no company witnesses): 120,000 AI cameras and no federal law to rein them in

By Zendoric · September 24, 2026.

A Senate judiciary subcommittee held on Wednesday the first congressional hearing on AI-powered video surveillance networks, titled "Always Watching: Flock's Nationwide AI Surveillance Network". It was convened by Senator Josh Hawley (Republican of Missouri), who chairs the Crime and Counterterrorism subcommittee and had already opened an investigation into Flock Safety, the leading company in the market for automatic license plate readers (ALPR: cameras that photograph and process vehicle license plates). As Hawley recounted during the session, he invited the top executives of Flock, Axon, Motorola and Verkada to testify. None showed up.

The figures discussed in the room give the measure of the problem. According to Alasdair Whitney, an attorney with the Institute for Justice, Flock operates more than 120,000 cameras across 49 states and records 20 billion vehicle photographs a month. The system does not stop at reading the plate: it builds what the company itself calls a "vehicle signature", combining make, model, color and details such as stickers, decals or accessories. Data is kept by default for 30 days before being automatically deleted, a window that civil rights organizations consider too long and conducive to building detailed profiles. There is, Whitney stressed, no federal law regulating any of this.

The hearing opened with the testimony of Lindsey Isaacs, a 23-year-old from Florida who spent 13 days in jail after police, relying on a mistaken identification of her car through Flock cameras, linked it to a fatal accident. Isaacs recounted that she spent nearly 86 hours in isolation, in an anti-suicide smock with broken Velcro that left her exposed. It is, as far as the available material allows one to judge, a case of algorithmic misidentification without sufficient human verification before an arrest, not a proven accusation against Flock by the Senate; The Hill notes that it requested comment from the company.

The most revealing point of the session, however, was not technical but a matter of data architecture. Several Democratic senators asked whether federal agencies, in particular Immigration and Customs Enforcement (ICE), have access to Flock's network. Senator Dick Durbin (Democrat of Illinois) recalled that his state prohibits by law the use of ALPR data in immigration investigations. Chad Marlow, of the American Civil Liberties Union (ACLU), answered with an uncomfortable precision: if the federal government presents a court order or an administrative subpoena, state law protects nothing, the data is handed over anyway. And if the state takes part in a data-sharing network among agencies, it is enough for a single officer to be willing to cooperate for ICE, or anyone else, to gain access through them. A state's law protects on paper; the architecture of the network decides in practice.

Here it is worth separating two things that coverage tends to conflate. License plate recognition technology is not a novelty of generative AI: it is mature computer vision, in use for more than a decade. What has changed is the scale and the centralization: a single private company operates a de facto national database, with no federal regulatory ceiling, with access rules that depend on commercial agreements between local police forces rather than on a uniform legal framework. The risk is not in the camera: it is in who decides who looks at what the camera sees, and with what safeguards before that data turns into an arrest.

Our reading is that this is exactly the kind of case in which evidence-based governance should prevail over panic and complacency alike. The ability to identify vehicles automatically can be, if well governed, a legitimate public safety tool: locating stolen cars or missing minors in minutes instead of days is a real benefit, not a hypothetical one. But the Isaacs case shows that without minimum requirements —mandatory human verification before an arrest, traceability of who consulted what data and why, uniform retention limits, a real firewall between local databases and federal agencies when the state has decided otherwise— the same technology produces concrete, verifiable harm, not an abstract risk ten years out.

In the short term, the problem is one of governance, not algorithms: Hawley noted that there are "many players" in this market and that Congress needs to know, in his words, "where this is taking us as a country"; Durbin, for his part, insisted that it falls to Congress to lead the balance between privacy and security, not to leave it in the hands of each company. Several jurisdictions are already canceling their contracts with these surveillance companies, and several lawmakers have introduced bills in recent months to restrict the use of Flock's cameras, although this year's legislative calendar itself works against them. That neither Flock nor its competitors turned up to testify is no minor detail: in a market without federal law, corporate silence in the face of public scrutiny is also a way of continuing to operate without rules. The window to set that framework, before the infrastructure becomes too big to regulate meaningfully, is closing faster than the debate.