Clearview AI data shows GAO's recent report on government use of facial recognition overlooked usage by five agencies, including IRS, TSA, and US Capitol Police
The mismatch matters because GAO is Congress's audit arm: if its census of government facial recognition use is incomplete, the baseline every oversight decision rests on is wrong, and the only party with a full picture is the vendor itself.
First-order effects
The IRS, TSA, and US Capitol Police are now publicly identified as Clearview users they apparently did not disclose to GAO, putting each directly in front of congressional questioners about who authorized the searches and on whose data.
GAO's just-published agency count is immediately shown to be an undercount, weakening the report's value as the definitive map of federal facial recognition use.
Second-order effects
GAO faces pressure to re-baseline its survey using vendor-side records rather than agency self-reporting, since the agencies themselves demonstrably don't know or won't say what their staff are running.
Clearview gains leverage from being the de facto record-keeper of government usage — a company with no formal federal contracts per the later CyberScoop reporting, yet holding the most complete accounting of which agencies search whose faces.
Third-order effects
If vendor logs keep proving more accurate than agency disclosures, federal oversight of biometric tools shifts from auditing agencies to subpoenaing vendors — an inversion with no established legal framework.
The pattern points toward facial recognition procurement outrunning accountability structures: usage spreads bottom-up through individual employees and units faster than any central inventory or ban regime can track.
The trend: Government facial recognition adoption is outrunning official inventories, leaving private vendors like Clearview AI holding better records of public-sector use than the auditors tasked with tracking it.
While 6 federally funded agencies used some type of facial recognition tech on BLM protesters, only 3 said they used the same tech to find suspected Capitol insurrectionists. “They have a lot of flexibility to hide what's really going on.” - @FoxCahn https://www.buzzfeednews.c…
@rajiinio
Deb Raji on x
ICYMI @USGAO just released a report on federal agency use of facial recognition. “...six (agencies)...using the technology to identify people...during the civil unrest, riots, or protests following death of George Floyd in May 2020” 🤔 h/t @jovialjoy https://www.gao.gov/...
@internetrebecca
Rebecca Williams on x
love this I wish GAO/IGs had more resources for thorough audits, they often rely on interviews where agency reps are unaware of “shadow IT” somehow here the agency knows a little, GAO knows more, Ryan knows the most it should make you think about oversight laws https://twitter.co…
@caro1inehaskins
Caroline Haskins on x
New: 5 federal agencies said they didn't use Clearview AI use in a new gov watchdog report on facial recognition. But...... this contradicts our previous reporting: https://www.buzzfeednews.com/ ...
@buzzfeedpol
@buzzfeedpol on x
In a government accountability report about facial recognition, five federal agencies said they didn't use Clearview AI. Data reviewed by BuzzFeed News suggests they should have. https://www.buzzfeednews.com/ ...
@rmac18
Ryan Mac🙃 on x
The GAO's report showing most federal agencies declaring they have no way of tracking the use of facial recognition software is worrying. The fed gov is saying it has no way of auditing if/how/when FRT is being deployed. And this will continue without federal regulation of FRT.
@rmac18
Ryan Mac🙃 on x
These five agencies that told GAO they didn't use Clearview: US Capitol Police, US Probation Office, Pentagon Force Protection Agency, TSA, and the IRS Criminal Investigation Division. Data we reviewed showed their employees using Clearview as of Feb 2020. https://www.buzzfeednew…
@rmac18
Ryan Mac🙃 on x
Yesterday the GAO published a report on federal use of facial recognition like Clearview AI. Except 5 agencies said they didn't use Clearview, when we've reviewed data showing that they had. It suggests the fed gov can't even track how FRT is being used. https://www.buzzfeednews.…