Log In Sign Up

Why Congress is probing police use of facial recognition technology

Lawmakers on Capitol Hill have opened a fresh line of inquiry into how law enforcement agencies deploy facial recognition systems, asking whether the technology is being used without proper oversight, transparent standards, or meaningful checks against error. Hearings held by the House Oversight Committee and Senate Judiciary panel this year have pulled in testimony from academics, civil liberties advocates, police chiefs, and the vendors who build the algorithms.

For readers in Sydney, Melbourne, or Perth, the discussion lands closer to home than it might first appear. Australia has its own long-running debate over the technology, from the federal Identity-Matching Services framework to the rollout in Service NSW outlets. The choices made in Canberra, and the way state forces such as NSW Police and Victoria Police have trialled matching tools, mirror the political fault lines now showing up in Washington.

What hangs in the balance is not just whether an algorithm can pick a face out of a crowd, but who decides when that capability is switched on, what evidence is admissible in court, and which communities end up over-represented in the watchlists.

The catalyst: misidentification cases spark bipartisan concern

Three high-profile wrongful arrest stories have done most of the heavy lifting in shifting the political mood. In each case, a Black man was detained after an algorithmic match later shown to be wrong. Lawmakers replayed the incidents frame by frame, underscoring how a probabilistic score can be treated as near-certainty by frontline officers under time pressure.

Researchers told the committee that peer-reviewed audits show commercial face-matching tools still produce higher error rates for women, the elderly, and people with darker skin. Vendors disputed the figures but few denied demographic disparities exist at all. Lawmakers from both parties flagged the gap between marketing claims and courtroom risk, and several raised the prospect of legislation requiring independent accuracy testing before procurement.

Federal agencies face scrutiny over warrant practices

A second strand has focused on the FBI, Customs and Border Protection, and Immigration and Customs Enforcement. Senators pressed officials on how often agents run a suspect photo through a facial recognition database, whether a warrant is obtained first, and what happens to images of people who are never charged.

Internal documents released during the hearings suggested millions of searches have been run with minimal documentation. Civil liberties lawyers argued this turns innocent travellers and bystanders into perpetual suspects. Agency leaders defended the practice as a fast triage tool, but several conceded that clearer logging rules and audit trails were overdue.

State and local police in a patchwork of policies

Outside the federal sphere the picture is murkier. Roughly half of US states have no statute governing police use of face-matching software, leaving individual departments to draft their own policies or accept vendor terms wholesale. Cities such as San Francisco, Boston, and Minneapolis have banned or sharply limited government use, while others continue to feed body-worn camera footage into real-time matching pilots.

Congressional staff have signalled interest in a baseline federal standard that would pre-empt the patchwork, requiring disclosure when an arrest or search relies in part on an algorithmic match. Industry groups have cautiously supported the idea, arguing that uniform rules would reduce the risk of successful constitutional challenges.

How Australia grapples with the same surveillance question

Closer to home, the Australian Human Rights Commission has raised comparable concerns about the national facial verification system behind Services Australia, passport renewals, and the federal driver licence database. The Identity-Matching Services Bill 2019, which allows agencies to share biometric data, passed with limited public debate and has since been criticised by legal scholars for its broad exemptions.

NSW Police admitted in 2021 that officers had trialled Clearview AI without formal approval, while Victoria Police ran a separate procurement that drew questions from the state Ombudsman. Australian Privacy Principle 3.3 already limits sensitive data collection to what is reasonably necessary, but the Office of the Australian Information Commissioner has called for clearer thresholds on real-time crowd scanning in public spaces.

Industry response and the push for regulation

Companies that sell face-matching tools are positioning themselves ahead of the rule-making. Several have released transparency reports, opened their models to external audit, and backed a federal accuracy standard modelled on National Institute of Standards and Technology benchmarks. Critics counter that self-regulation is not enough when procurement contracts are lucrative and penalties for misuse are low.

A handful of vendors have refused to sell to certain agencies or pulled products from markets without independent oversight. The technology itself is no longer the central question. The central question is who controls it, under what authority, and with what recourse for the person standing in front of the camera.

Practical safeguards policymakers could adopt

Share this story with a friend who follows surveillance policy, or subscribe to the channel for weekly updates on technology regulation in Australia and beyond.