ID.me identity verification

ID.me is a remote identity check run by the private company of the same name. A claimant uploads a government-ID photo and a live video selfie, and facial-recognition software compares the two. During the pandemic, state unemployment agencies placed the check in front of benefit payment, so a failed match held up the claim.

How it was used

ID.me signed its first state contract, with Florida, in June 2020. Within about a year, it held contracts with at least 25 state unemployment agencies, for nearly $45 million. It served identity checks for roughly 30 state governments and 10 federal agencies. The IRS alone spent about $86 million on licenses.

A claimant needed a smartphone or a computer with a camera. The software's pass or fail decided whether the claim went ahead. Roughly 10 to 15 percent of applicants were sent on to a live video interview with an ID.me trusted referee.

A Department of Labor Inspector General audit, reported on March 31, 2023, found that 24 of 53 state workforce agencies, 45 percent, used a facial-recognition identity contractor. They drew on 10 different vendors.

The wait for a person

House investigators found that the wait for the video interview averaged nearly 10 hours in North Dakota in April 2021. It averaged nearly 6 hours in Washington, and over 4 hours in 14 of 21 states.

New Jersey, the only state then offering an in-person alternative, averaged about 6 minutes. ID.me removed appointment scheduling, which was judged to be "hindering efficiency." The sources do not say who judged it so. It told the IRS that waits were "about 2 hours," while its own data showed averages over 4 hours across most states.

A check that records no denial

ID.me's check does not score anyone or decide eligibility. It is a fraud-control checkpoint placed in front of the claim. A claimant who does not finish an hours-long wait has not been denied. There is no decision to appeal, no reason code, and no letter.

The US Department of Labor does not collect or report how many workers are denied benefits because they could not verify their identity. Where verification comes before filing, blocked workers are not counted as denied claims at all. The National Employment Law Project, an advocacy group, documented this.

The gap here is not a bad face match. It is a denial in practice with no denial to appeal, and no count of who was lost.

Who completed verification

Oregon studied one month of regular unemployment claims sent to ID.me, 10,656 in all. By a November 2021 snapshot, 49.37 percent had completed verification. Eventually 71.06 percent did, so roughly 29 percent never completed.

Completion varied by group. It was 53.44 percent for White claimants, 41.59 percent for African American claimants, and 42.05 percent for Native Hawaiian or Pacific Islander claimants. It was 49.65 percent for English-language filers and 34.48 percent for Spanish-language filers.

Claimants aged 20 and under completed at 28.49 percent. Telephone filers completed at 38.25 percent, against 52.10 percent for internet filers. The lowest weekly benefit band completed at 39.52 percent, against 60.32 percent for the highest.

Oregon stressed that the study showed differences in completion and "did not show causation." Its outreach to 130 non-completers found that some faced a technology barrier or found the process confusing. Others had returned to work (22 percent) or had not finished by the snapshot.

Who gains and who pays

The agencies see a real, measurable benefit. Of 24 surveyed agencies, 22 (92 percent) reported that the technology reduced improper payments. Those are the agencies' own reports, not independently audited.

The cost, the wrongful lockout of eligible claimants, goes unmeasured. The case file says it falls hardest on filers with poor broadband, filers who do not speak English, younger filers, and those with the lowest benefits. Oregon's completion gaps measure friction, not errors, but the case file says they plainly show who bears the cost.

ID.me also had a reason to overstate the fraud it sells protection against. Its chief executive claimed that as much as $400 billion was lost to pandemic unemployment fraud. The Inspector General estimated $45.7 billion in potential fraud.

What changed afterward

On October 29, 2021, Oregon paused the automatic routing of all regular unemployment claims to ID.me. It did so to study disparate impacts, meaning whether the check affected groups differently. The case file calls this the one safeguard the sources show being used, and the cheapest in this area of public benefits.

After bipartisan backlash, the IRS and the Treasury dropped their mandatory facial-recognition requirement in February 2022. ID.me made facial recognition optional across government agencies. Massachusetts became the first state to follow, though there the technology had already been optional.

In July 2023, the Department of Labor launched a public alternative. It pairs Login.gov, run by the General Services Administration, with in-person checks at the US Postal Service. A 2024 review by the Government Accountability Office found that federal agencies had leaned heavily on commercial identity vendors. It found that Login.gov had only recently added one-to-one face matching that meets the guidance of the National Institute of Standards and Technology.

ID.me and similar services still checked unemployment claims in a large share of states into 2025. In 2026, the IRS proposed allowing ID.me to keep taxpayer biometric data for up to 36 months after an account is deleted.

What this network is drawn from

This network follows the identity-check pattern in the public record of the ID.me case. It is not a reconstruction of ID.me's system, its algorithm, or any state's setup.

It shows ID.me's check, what the check receives, and the wait for a trusted referee. It also shows the referees and state staff, the verification outcome record, the face database and ID.me's scan storage, and the outside reviewers. Claimants themselves are outside the network.

What the available tools can and cannot address

Service means the useful work ID.me's check does for the agencies: confirming real claimants and stopping improper payments.

A failure pathway is a link between two parts of the network where a mistake made by one part can be passed on to the other. Closing a pathway means mistakes stop passing along it. The work it carries goes on.

This case has a budget of 14 units. Explore (No Targets) sets no targets.

Under Service Targets Only, the targets can be met within the budget. No single tool meets them. The cheapest way costs 5 units and uses two tools: Escalate checks and Vet connections.

Escalate checks has the trusted referees and state staff check more closely when monitoring flags trouble. Here it stops mistakes passing along Failed match to a trusted referee. Vet connections lets one system read or copy another's records only where that is explicitly granted. Here it closes three pathways: Images and fraud flags to ID.me, Earlier flags send claimants back, and Facial scans kept by ID.me.

Under Service and Safety Targets, the targets can also be met. That level also asks you to close every failure pathway, among other targets. The cheapest way costs 8 units and uses three tools: Escalate checks, Store less data, and Vet connections.

Under All Governance Targets, the targets can also be met, with less room to spare. The cheapest way costs 11 units and uses four tools: Understand the system, Escalate checks, Store less data, and Vet connections. At this level, Understand the system costs 4 units and makes Vet connections 1 unit cheaper.

At that level, three tools act less strongly unless Understand the system is also in use: Check copied records, Store less data, and Vet connections. Without it, Vet connections closes none of its pathways. Other combinations of tools also meet the targets at every level that sets them.

More is not better here. Using every tool at its strongest setting closes every failure pathway and costs 33 units, more than twice the budget. It also leaves ID.me's check doing too little useful work, so it misses the targets at every level that sets them.