Clearview AI Algorithm Built on Photos Scraped from Social Media Profiles without Consent

Face-matching algorithm by Clearview AI was built using scraped images from social media sites such as Instagram and Facebook without user consent, violating social media site policies, and allegedly privacy regulations.

Risk classification

• Primary risk domain: 4 Malicious actors
• Primary risk subdomain: 4.1 Disinformation, surveillance, and influence at scale

The system is used to conduct large-scale, automated mass surveillance by matching faces against billions of scraped photos, threatening public anonymity.

Additional risk subdomains

• 1.3 Unequal performance across groups: The facial recognition technology exhibits higher error and misidentification rates for people of color, exacerbating systemic biases in law enforcement.
• 2.1 Compromise of privacy by obtaining, leaking or correctly inferring sensitive information: The AI extracts unique biometric faceprints and links public photos to reveal private identities and online footprints without user consent.

Causal factors

• Entity: AI
• Intent: Intentional
• Timing: Post-deployment

The privacy risks and mass surveillance capabilities are driven by the intentional deployment and operation of Clearview's AI facial recognition system.

EU AI Act risk tier

• Risk tier: 2 High Risk

Risk Level 2: High Risk. The reports describe Clearview AI as a facial recognition tool used by law enforcement agencies, which falls under 'Law enforcement applications, including facial recognition' as a High Risk AI system.

AI system and alleged parties

• AI system: Clearview AI
• AI purpose: Face Recognition; Identification
• Behaviour type: Tool
• Alleged developer: Clearview AI
• Alleged deployer: Clearview AI
• Alleged harmed parties: social media users, Instagram users, Facebook users

Harm severity

Severity ratings are based on this scale.

Highest direct severity in any category: Severe. Severity is scored from Negligible to Catastrophic in each harm category, for harm the reports describe as caused directly or indirectly by the AI system.

• Physical: direct Negligible, indirect Negligible
• Infrastructure: direct Negligible, indirect Negligible
• Property: direct Negligible, indirect Negligible
• Financial: direct Substantial, indirect Negligible
• Environmental: direct Negligible, indirect Negligible
• Malicious content: direct Negligible, indirect Negligible
• Differential treatment: direct Substantial, indirect Substantial
• Civil rights: direct Substantial, indirect Substantial
• Democracy: direct Minor, indirect Substantial
• Privacy: direct Severe, indirect Severe
• Psychological: direct Minor, indirect Negligible
• Epistemic: direct Minor, indirect Negligible
• Child sexual exploitation and abuse: direct Negligible, indirect Negligible

Financial

Reported: The report explicitly describes a 20 million euro penalty imposed on Clearview AI by the French data protection authority (CNIL).

Directly caused: Clearview AI was fined 20 million euros for collecting and using data on individuals in France without a legal basis.

Inferred additional harm: Unquantified legal defense costs and compliance expenses incurred by Clearview AI and the platforms issuing cease-and-desist letters.

Differential treatment

Reported: The report explicitly describes how the technology disproportionately harms and targets immigrants, people of color, and political activists.

Directly caused: The system enhances law enforcement's ability to monitor and target specific groups, including Black Lives Matter activists and immigrants.

Indirectly caused: The technology carries a higher risk of misidentifying people of color due to inherent algorithmic biases.

Inferred additional harm: An increased rate of wrongful investigations or arrests of marginalized groups due to biased facial recognition matches.

Civil rights

Reported: The report explicitly describes violations of constitutional privacy rights and the chilling of protected political speech and protest.

Directly caused: The lawsuit alleges that Clearview's mass surveillance technology violates the privacy rights of Californians and chills free speech.

Indirectly caused: The widespread adoption of the tool by law enforcement undermines the basic civil liberty of public anonymity.

Inferred additional harm: Long-term erosion of the right to protest and freedom of association due to pervasive, clandestine tracking.

Democracy

Reported: The report explicitly describes the erosion of democratic norms through the circumvention of local government bans on facial recognition.

Directly caused: Clearview AI bypassed municipal bans in cities like San Francisco and Oakland by continuing to sell its technology to local law enforcement.

Indirectly caused: The technology facilitates a transition toward an unaccountable mass surveillance state.

Privacy

Reported: The report explicitly describes massive privacy violations through unauthorized scraping and biometric profiling.

Directly caused: Clearview AI scraped over 3 billion photos of individuals from public websites and extracted unique biometric faceprints without consent.

Indirectly caused: The company stores scraped images and metadata indefinitely, even after the original photos are deleted or made private by users.

Inferred additional harm: Millions of individuals globally have had their biometric privacy permanently compromised without recourse.

Psychological

Reported: The report explicitly describes psychological distress and a chilling effect on political speech caused by the threat of clandestine surveillance.

Directly caused: Activists and immigrants report feeling threatened and experiencing a chilling effect on their protected political speech and protest actions.

Inferred additional harm: Widespread anxiety and distress among targeted communities who feel they are constantly subjected to an invasive police lineup.

Epistemic

Reported: The report explicitly describes misleading marketing claims regarding the system's accuracy.

Directly caused: Clearview AI falsely claimed its technology was '100% accurate' across all demographic groups based on a self-conducted test mimicking ACLU methodology.

People affected

• Occurrences reported: 1
• People reportedly harmed: 3000000000
• People reportedly exposed: 3000000000

Potential causes

Management

• Prioritizing Growth Over Safety: Stealth-mode expansion focused on viral adoption by law enforcement.
• Disregard for Platform Rules: Management ignored cease-and-desist letters from major tech platforms.
• Lack of Transparency: Operating in secrecy with fake names and hidden business addresses.

Technology

• Algorithmic Racial Bias: Studies showed facial-recognition performs poorly on people of color.
• Testing Environment Discrepancy: System tested on clear headshots instead of grainy real-world footage.
• High Doppelganger Risk: Large databases increase risk of false matches and wrongful arrests.

Data Inputs

• Unauthorized Web Scraping: Scraped billions of photos from social media violating terms of service.
• Lack of User Consent: Biometric profiles created without knowledge or consent of individuals.
• Indefinite Data Storage: Images and metadata stored even after being deleted from original sites.

Human Factors

• Overreliance by Law Enforcement: Officers adopted the tool rapidly without understanding its limitations.
• Misleading Marketing Claims: Clearview falsely claimed 100 percent accuracy using misleading studies.
• Lack of Technical Expertise: Review panel lacked expertise in validating facial recognition accuracy.

Process and Methods

• Insufficient Independent Testing: Accuracy claims were not vetted by independent bodies like NIST.
• Flagging of Anomalous Searches: System allowed manual manipulation of search results to block media scrutiny.
• Undisclosed Search Storing: Uploaded sensitive photos were stored on servers without officer awareness.

Regulatory Environment

• Absence of Federal Privacy Law: Lack of US federal rules allowed unregulated mass surveillance proliferation.
• Outdated Legal Statutes: Pre-internet laws like CFAA are inadequate to prevent modern web scraping.
• Local Ban Circumvention: Software sold to police departments in cities where the tech was banned.

Information quality

• Classification confidence: High
• Reason for confidence: The reports provide consistent, detailed, and well-documented accounts of Clearview AI's scraping practices, database size, law enforcement usage, and the resulting legal and regulatory challenges across multiple jurisdictions.
• Ambiguities identified: Whether the system has led to specific wrongful arrests is disputed, with Clearview claiming no known instances while critics point to general facial recognition failures.
• Alternative interpretations: Clearview AI argues its scraping is protected under the First Amendment as public information gathering, presenting an alternative legal interpretation of its actions.

Clearview AI's unauthorized scraping of over three billion images to create a biometric database for law enforcement represents a substantial threat to societal stability and civil liberties. The deployment of this unvetted mass surveillance tool by federal agencies undermines privacy rights, risks systemic racial bias, and has caused significant regulatory and legal friction internationally.

• Overall national security impact: Substantial
• Scope: Multiple nations
• Primary target: United States
• Other affected: Canada, European Union
• Alleged perpetrator: Clearview AI

Threat characteristics

• Imminence: Long-term. Represents an ongoing, systemic capability and privacy concern rather than an active, fast-moving crisis.
• Autonomy: Human-controlled. The tool assists law enforcement by matching images, but human operators make all final decisions regarding investigations and arrests.
• Novelty: First-of-its-kind. Established a highly controversial precedent of mass-scraping public internet data to build a private biometric database for state surveillance.

Impact by dimension

• Physical security: Negligible. No kinetic threat, critical infrastructure compromise, or physical safety risks were reported in this incident.
• Information security: Minor. Represents an unvetted mass surveillance capability deployed to law enforcement, but does not involve adversarial information warfare or compromise of classified state intelligence.
• Sovereignty: Minor. Bypassed local municipal bans on facial recognition and caused regulatory friction with international allies who declared the technology illegal.
• Economic security: Negligible. Financial impacts are limited to regulatory fines and legal costs for a single private entity, with no systemic threat to economic security.
• Societal stability: Substantial. Implements a massive, unvetted mass surveillance capability violating the privacy rights of billions, chilling political speech, and carrying high risks of racial bias in law enforcement.