CBP Bought Clearview AI a Week After Congress Tried to Ban It

TL;DR: U.S. Customs and Border Protection signed a one-year, $225,000 contract for 15 Clearview AI licenses at its National Targeting Center. The system searches a database of 60 billion+ images scraped from social media, news sites, and anywhere else your face appears online. CBP calls it “tactical targeting” and “counter-network analysis.” The contract doesn’t say whether agents can search U.S. citizens’ faces, what images they can upload, or how long results are kept. This landed the same week Congress introduced the ICE Out of Our Faces Act to ban the technology entirely. One branch of government is buying. Another is trying to outlaw it.

What changed (September 9, 2026): The Seventh Circuit vacated Clearview AI's $51.75 million nationwide class-action settlement on July 13, 2026. The court held the deal was structurally defective because no named plaintiff adequately represented the nationwide class separate from the Illinois, California, New York, and Virginia subclasses. Clearview still owes the plaintiffs a remedy under BIPA and California, New York, and Virginia privacy law; the case is back in the district court and the equity-based settlement is no longer on the table as approved. The legal pressure on the same 60-billion-image database CBP is now buying into got weaker, not stronger, in the two months before the contract goes live.

The Contract

The deal went public on USASpending.gov in early February 2026 under contract number 70CMSD25P00000111. Here’s what’s in it ^[1]:

$225,000 for one year of Clearview AI access
15 software licenses for CBP analysts
Deployed at the National Targeting Center’s intelligence division
Extended to Border Patrol’s headquarters intelligence division
Starts September 2026
Purpose: “tactical targeting” and “strategic counter-network analysis”

CBP explained the investment would let analysts “more effectively and efficiently identify, target, screen, and interdict inbound and outbound passengers who pose a threat to national security” ^[1].

This isn’t CBP’s first encounter with Clearview. According to DHS’s own AI inventory, CBP began piloting the technology in 2025 ^[1]. The new contract formalizes what was already happening quietly.

What CBP Just Bought Access To

Clearview AI scraped over 60 billion images from the internet. Facebook. Instagram. LinkedIn. Dating apps. News sites. Your vacation photos. Your kid’s birthday party that someone posted publicly. All of it vacuumed into a searchable facial recognition database ^[2].

Upload a photo. Get matches. Get links to where those images originally appeared online. That’s the product.

Nobody in those 60 billion photos consented. No social media platform authorized the scraping. Clearview did it anyway, got sued, got fined across multiple countries, and kept selling to governments.

The company has racked up enforcement actions and fines across the globe:

Netherlands: €30.5 million fine (September 2024) ^[3]
France: €20 million fine (2022) ^[4]
Italy: €20 million fine ^[4]
Greece: €20 million fine ^[4]
UK: £7.5 million penalty from the ICO ^[5]
Australia: Found to have breached privacy law, though the regulator later dropped enforcement ^[6]

Total European fines alone: over €90 million. Clearview’s response to all of it has been, essentially, that foreign regulators can’t reach a U.S. company. They haven’t paid ^[3].

What the Contract Doesn’t Say

The contract contains nondisclosure requirements for contractors handling biometric data. It does not contain answers to the questions that actually matter ^[1][2]:

Can agents search U.S. citizens’ faces? Not specified.
What images can agents upload? Not specified.
How long are search results retained? Not specified.
What oversight exists for individual searches? Not specified.
Are there audit logs? Not specified.

DHS’s AI inventory includes a carefully worded assurance: “No enforcement action is taken based solely on the leads generated by this tool. All potential identifications undergo thorough investigation and validation” ^[1].

That sounds reassuring until you think about it. “Not based solely on” means Clearview results can be a factor in enforcement actions. And “thorough investigation and validation” is whatever the agent doing the search says it is.

The Accuracy Problem

NIST testing on facial recognition systems shows they perform well on high-quality, controlled photos: studio headshots, passport images, that sort of thing. But error rates spike for uncontrolled images: street photos, social media snapshots, anything taken in the real world. Sometimes those error rates exceed 20% ^[2].

NIST’s recommendation: use these systems only for “investigative” purposes (generating ranked lists of possible matches) because automatic matches produce false positives ^[2].

CBP’s contract calls this “tactical targeting.” That language suggests real-time or near-real-time identification in operational settings. Not an analyst sitting at a desk reviewing a ranked list. That’s a gap between how NIST says these systems should be used and how CBP appears to be deploying them.

DHS’s Facial Recognition Shopping Spree

CBP’s $225,000 Clearview contract isn’t happening in isolation. It’s part of a DHS-wide expansion of biometric surveillance:

ICE signed a $9.2 million Clearview AI contract in September 2025, nearly 40 times larger than CBP’s deal ^[7]
DHS maintains a database of 1.2 billion face images through NEC’s Mobile Fortify system, used by ICE agents in the field ^[8]
The Pentagon appointed a former Clearview AI advisory board member (Owen West) to run its $2 billion Defense Innovation Unit ^[9]
ICE agents carry smartphones loaded with facial recognition that can scan faces during encounters on the street ^[10]

Albert Fox Cahn of the Surveillance Technology Oversight Project put it bluntly: “The use of Clearview AI by federal agencies is essentially mass surveillance laundered through a private company” ^[11].

Congress Pushes Back: Sort Of

On February 5, 2026 (days before the CBP contract went public), Senators Edward Markey, Jeff Merkley, and Ron Wyden, along with Representative Pramila Jayapal, introduced the ICE Out of Our Faces Act ^[12].

The bill would:

Ban ICE and CBP from buying or using facial recognition technology
Require deletion of all biometric data already collected
Allow individuals and state attorneys general to sue for civil penalties

Senator Merkley: “Without oversight, this technology is dangerous in the hands of any government” ^[1].

The bill has backing from the ACLU, EFF, EPIC, Fight for the Future, Human Rights First, and Access Now. It’s also the most aggressive federal facial recognition ban ever proposed for immigration enforcement.

It has zero chance of passing the current Congress.

Which means CBP’s Clearview contract will go live in September, and the legislative ban will collect dust in committee. That’s not cynicism. That’s arithmetic.

What You Can Do

Limit Your Public Face Data

Clearview scrapes publicly accessible images. Set your social media profiles to private. Remove old photos from public-facing platforms. Every photo you leave public is another data point in a 60-billion-image database.

Support the Legal Fight

The ACLU’s facial recognition campaign and EFF’s About Face project track cases and push for bans. Your state may have pending biometric privacy legislation.

Know the Opt-Out (It’s Barely an Opt-Out)

Clearview AI offers a privacy request form where you can ask to be excluded. But you have to provide a photo of yourself to request removal, giving them the very data you want deleted. And it doesn’t stop law enforcement from searching you.

Contact Your Representatives

The ICE Out of Our Faces Act needs co-sponsors. Call your senators and representative. Reference the bill by name. Tell them you oppose CBP’s use of scraped biometric data for immigration enforcement.

Update (September 9, 2026)

On July 13, 2026, the U.S. Court of Appeals for the Seventh Circuit vacated the $51.75 million class-action settlement that had been the centerpiece of the multi-year consumer privacy litigation against Clearview AI. The settlement, finalized in March 2025 and approved by a district court, would have given class members a 23 percent equity stake in Clearview (or a cash fallback), with allocation skewed toward Illinois subclass members (10 shares) over California, New York, and Virginia subclasses (5 shares each) and the nationwide class (1 share). The Seventh Circuit's opinion in In re Clearview AI, Inc. Consumer Privacy Litigation, No. 25-1673, found a "key procedural problem": none of the named plaintiffs was positioned to represent the nationwide class separate from the state subclasses, so the court could not confirm absent class members were fairly represented. The court rejected Clearview's arguments that the deal lacked injunctive relief and that the equity-based structure was too uncertain, calling both "inherent" in such settlements. The case is back in the district court for further proceedings.

The settlement was the legal pressure point most likely to make Clearview change its business practices. The vacated deal means the underlying claims under Illinois's Biometric Information Privacy Act and the California, New York, and Virginia privacy statutes all remain live. Clearview's defense posture is also weaker in light of the Seventh Circuit's separate April 1, 2026 decision in Clay v. Recon Protective Services holding that the 2024 BIPA amendment limiting damages to a single recovery applies retroactively to pending cases, a ruling that cuts both ways for Clearview: it caps per-scan damages, but it does not eliminate them.

The practical effect for CBP's $225,000 contract: the database of 60 billion+ scraped images that the September 2026 deployment at the National Targeting Center will search is the same database that the consumer litigation has been trying to constrain since 2020. The legal pressure on that database got weaker, not stronger, in the two months before the contract goes live. Clearview was named to the Inc. 5000 list for the third consecutive year in August 2026, an indication that the company is still expanding even as the appellate litigation resets.

Sources for this update: Weissman v. Clearview AI, Inc., No. 25-1673 (7th Cir. July 13, 2026); Duane Morris. "Seventh Circuit Undoes Novel Privacy Class Settlement Due to Lack of Separate Representatives for Nationwide Class and State Sub-Classes." July 14, 2026; National Law Review. "Seventh Circuit Holds Amendment Decreasing BIPA Exposure Applies Retroactively." April 2026; Clearview AI Press Room.