AI and Privacy: How Your Face Becomes Data Without Your Consent

Social media scraping

Billions of publicly posted photos have been scraped from Facebook, Instagram, Twitter, LinkedIn, and smaller platforms to build facial recognition training datasets. Clearview AI scraped over 30 billion images from the public web and sold search access to law enforcement. Several academic datasets — later retracted — were built the same way. Every public photo you post is a potential training data point.

Photo-tagging systems

When you tag someone in a photo on a major platform, you are contributing a labelled training sample: this face belongs to this named person. Facebook's DeepFace system was trained largely on user-generated photo tags. These systems build a faceprint — a mathematical embedding of your facial geometry — that persists even if you delete the photo.

Public surveillance cameras

Cameras in retail stores, transport hubs, stadiums, office lobbies, and public streets increasingly feed into facial recognition systems. In many jurisdictions this is legal and requires no notice. Your face may be matched against a database every time you walk through a shopping centre — with no way of knowing and no log you can access.

Consumer devices

Smartphone face unlock systems, photo organisation apps (Google Photos, Apple Photos), and smart doorbells all use on-device or cloud-based facial recognition. Even 'on-device' processing often shares anonymised embeddings with the manufacturer for model improvement. The permissions granted during setup are rarely read in full.

Data broker aggregation

Data brokers combine photos from public records, social media, and purchased datasets with identity information — name, address, employer, relatives — to create rich profiles. These profiles are sold to advertisers, insurers, employers, landlords, and government agencies. Face data is increasingly included as a biometric identifier alongside fingerprints and voice recordings.

Technology companies

Meta, Google, Apple, and Amazon all operate facial recognition systems — for photo organisation, device unlock, and increasingly for commerce. Their privacy policies disclose data collection in broad terms but rarely explain how faceprints are retained, shared, or used for model training after the named feature is used.

Facial recognition vendors

Companies like Clearview AI, NEC, Idemia, and Rank One sell search and identification services. Law enforcement is the primary customer, but insurers, financial institutions, and private investigators also purchase access. These companies operate largely outside public scrutiny and face minimal regulation in most countries.

Retailers and hospitality

Major retailers use facial recognition to flag known shoplifters, match VIP customers, and track foot traffic patterns. Some casinos match faces against banned-player lists. Customers are rarely notified and have no meaningful mechanism to opt out — short of not entering the premises.

Governments and law enforcement

Police forces in dozens of countries use facial recognition to identify suspects from CCTV footage, passport databases, and in some cases social media. Error rates vary dramatically by demographic group — a 2020 NIST study found false positive rates up to 100x higher for certain demographic groups compared to others.

Advertisers and analytics firms

Emotion recognition — inferring mood from facial expressions — is used in advertising research, job interview screening tools, and customer experience analytics. Several countries have moved to restrict this application, but it remains widespread.

Faceprints are permanent even after deletion

Deleting a photo from a social media platform does not delete the faceprint extracted from it. The mathematical embedding has already been computed and stored — or incorporated into model weights — before the deletion occurred. Most platforms' data deletion policies cover stored images, not derived biometric data.

Cross-database matching

A faceprint from a social media profile can be matched against CCTV footage, a passport database, or a retail surveillance system — even if those databases were built independently. The geometry of your face is a universal identifier that links records across systems that were never intended to be connected.

Secondary sale without notice

In many jurisdictions, data collected under one privacy policy can be transferred or sold to a third party after an acquisition, bankruptcy, or policy change. The company that built the facial recognition product you used five years ago may no longer exist — but the data it collected may be in dozens of other hands.

Use in training future models

Photos and faceprints collected today are used to train the next generation of recognition models. This means the impact of current collection extends decades into the future. A more accurate model trained on your data may identify you in contexts where today's model would fail.

Aggregation with other biometric data

Facial data is increasingly combined with voice recordings, gait analysis, and behavioural biometrics to create multimodal profiles. A profile that includes your face, voice, and walking pattern is significantly harder to avoid than any single identifier — and is already used in some border control and financial fraud detection systems.

EU / UK — GDPR

Biometric data is a 'special category' under GDPR, requiring explicit consent for processing. You have the right to access data held about you, correct inaccuracies, and request deletion (the 'right to be forgotten'). You can also object to processing based on legitimate interests. File complaints with your national Data Protection Authority — in the UK, the ICO; in Ireland, the DPC. GDPR has teeth: fines can reach 4% of global annual turnover.

Illinois — BIPA

The Biometric Information Privacy Act is the strongest facial recognition law in the US. It requires written consent before collecting biometric identifiers, prohibits sale of biometric data, and creates a private right of action — meaning individuals can sue without proving harm. Clearview AI paid a settlement under BIPA. If you are an Illinois resident, this applies to you regardless of where the collecting company is based.

California — CCPA / CPRA

California residents can request disclosure of what biometric data is collected, opt out of its sale, and request deletion. The California Privacy Rights Act extended these rights and created the California Privacy Protection Agency to enforce them. Coverage is broad but enforcement has been slower than under GDPR.

Most other US states

Outside Illinois and California, federal US law has no specific biometric privacy protections. Several states — Texas, Washington, New York — have passed or are passing legislation, but enforcement varies. In most US states, a company can legally collect and sell your faceprint with only a terms-of-service disclosure.

How to exercise your rights

Submit a Subject Access Request (SAR) to any company you believe holds your biometric data — they are legally required to respond within 30 days under GDPR and 45 days under CCPA. If they fail to respond or refuse without justification, file a complaint with your regulator. For Clearview AI specifically, a deletion request form is available at their website — required by multiple legal settlements.

Restrict public photo access

Set social media profiles to private. Every public photo you post is a potential scraping target. This does not remove data already collected, but it stops the accumulation of new data points. Use distinct photos across platforms to limit cross-platform matching.

Disable photo tagging and facial recognition features

On Facebook: Settings → Privacy → Profile and Tagging → disable face recognition. On Google Photos: turn off face grouping under Explore settings. On iPhone: disable Face ID sharing under Privacy settings. These platform-level features generate labelled faceprints at scale.

Submit deletion requests to data brokers and facial recognition vendors

Clearview AI, PimEyes, and similar services offer opt-out or deletion mechanisms — often required by legal settlements. Submit requests to each. Also submit data deletion requests to the major social platforms under GDPR or CCPA. These requests cover stored biometric data, not just photos.

Run a face search on yourself

Before you can request removal, you need to know where your face appears. Run your profile photos through a face search engine to find which sites are using your image — then submit removal requests to those sites directly or via Google's image removal tool.

Be aware of physical surveillance limits

In-store and in-public facial recognition is largely outside your control in most jurisdictions. What you can do: be aware of which retailers and venues use it (several have disclosed this in privacy notices or through press coverage), avoid smart doorbell camera coverage where possible, and support legislative efforts to regulate public biometric surveillance in your region.