← Back to database

Predaxia Research

Clearview AI

Clearview AI Inc. · United States

ActiveBIPA settlement 2022Multiple EU fines
5

Confidence 5/5

VendorClearview AI Inc.
Country of originUnited States (New York, NY)
Founded2017
FoundersHoan Ton-That, Richard Schwartz
Current statusActive. Restricted to US government and law-enforcement customers (post-BIPA settlement May 2022)
US Entity ListNo
DatasetClaims 30 billion+ images scraped from public internet
Major regulatory actionsUK ICO, French CNIL, Italian Garante, Greek HDPA, Australian OAIC, Canadian OPC, Hamburg DPA
Predaxia GlossaryWhat it means in practicePlain-English explainer · Threat model · Action steps

Technical capabilities

Clearview AI operates a facial-recognition search engine built on a database of more than 30 billion images, scraped without consent from public sources including Facebook, LinkedIn, Twitter, Venmo, YouTube and news websites. Given a probe image, the platform returns matches with links to the source pages, enabling cross-platform identity correlation.

The vendor states accuracy of 98.6 percent against the National Institute of Standards and Technology Face Recognition Vendor Test (NIST FRVT) benchmark. Independent testing has demonstrated meaningful degradation on darker-skinned and female faces, consistent with the broader NIST findings on demographic differential in face-recognition algorithms (NIST Interagency Report 8280, December 2019).

Documented use

Clearview’s customer base has been disclosed through a combination of news reporting, court filings, and FOIA disclosures. Documented users include:

  • US Immigration and Customs Enforcement (ICE), US Federal Bureau of Investigation, US Department of Homeland Security, US Department of Justice components.
  • More than 600 US law-enforcement agencies at state and local levels, per the New York Times disclosure (January 18, 2020).
  • Ukrainian government, used to identify Russian soldiers and casualties since March 2022 (per company statements and Reuters reporting).
  • Past trials at Toronto Police Service (terminated 2020), Royal Canadian Mounted Police (terminated 2021), London Metropolitan Police, French DGSI and others, generally followed by termination after public exposure.

The American Civil Liberties Union (ACLU) settlement under the Illinois Biometric Information Privacy Act (BIPA), reached May 2022, prohibits Clearview from selling its database to most private businesses in the United States. The restriction does not apply to government customers.

Customer states

Clearview formally states that its service is now restricted to US government and law-enforcement customers since the BIPA settlement. Pre-settlement disclosures identified customers in dozens of countries including the United States, Canada, the United Kingdom, France, Italy, Sweden, Finland, Australia and the United Arab Emirates. Several of those deployments have since been terminated under regulatory pressure.

Legal and sanctions status

  • Not on the US Department of Commerce Entity List.
  • Not designated by US OFAC.
  • Illinois BIPA settlement (ACLU v. Clearview AI), May 9, 2022: nationwide ban on most private-business sales.
  • UK Information Commissioner: penalty notice GBP 7.5 million (May 23, 2022), order to delete UK resident data and stop processing.
  • French CNIL: penalty EUR 20 million (October 17, 2022), repeated daily penalty of EUR 5.2 million in May 2023 for non-compliance.
  • Italian Garante: penalty EUR 20 million (February 10, 2022).
  • Greek HDPA: penalty EUR 20 million (July 2022).
  • Australian Office of the Australian Information Commissioner: covert collection found unlawful (November 3, 2021), ordered destruction of Australian-resident data.
  • Canadian OPC and provincial commissioners: found unlawful surveillance, ordered cessation (February 3, 2021).
  • Hamburg DPA Germany: compliance order (January 27, 2021).

Technical countermeasures

  • Minimize public face exposure: review and reduce publicly indexed photos on LinkedIn, Facebook, Instagram and any platform allowing public scraping. Clearview’s database is built from previously-public material.
  • Adversarial-style imagery: research projects like Fawkes (University of Chicago SAND Lab) provide cloaking transforms for personal photos. Effectiveness against current Clearview models has limits but reduces baseline recognition.
  • Right to access and erasure: residents of EU, UK, California (CCPA/CPRA), Illinois, Texas, Virginia and increasing US states can request access, deletion, and opt-out from Clearview under applicable laws.
  • Avoid biometric self-tagging: do not opt into platform face-tagging features. Avoid publishing high-resolution frontal portraits with metadata intact.
  • Mask or sunglasses for protests: where lawful. Face occlusion is one of the simplest practical defenses against deployment-stage recognition.
For at-risk individuals. Clearview operates on data you may have published years before the company existed. Even if you delete your accounts today, your past public uploads may already be in its database. Submit data-deletion requests under applicable privacy law and assume residual exposure.

Sources

  1. The New York Times, The Secretive Company That Might End Privacy as We Know It (January 18, 2020)
  2. ACLU, Clearview AI BIPA settlement (May 9, 2022)
  3. UK ICO, Clearview AI penalty notice (May 23, 2022)
  4. CNIL, Délibération SAN-2022-019 against Clearview AI (October 17, 2022)
  5. Garante, Clearview AI sanction (February 10, 2022)
  6. NIST, Interagency Report 8280: Demographic Effects in Face Recognition (December 2019)
  7. Reuters, Ukraine uses Clearview AI in war effort (March 14, 2022)

Update log

April 19, 2026: Entry created. Sourced from the New York Times original disclosure (January 2020), regulatory decisions by UK ICO, French CNIL, Italian Garante and Greek HDPA, the ACLU BIPA settlement, NIST FRVT demographic research, and Reuters reporting on the Ukraine deployment.

There’s no perfect setup. Anyone selling you perfect is selling fear. The goal is simple: make yourself a harder target than the person next to you.