Who Owns Your Digital Self? The Emerging Market for Human Faces and Voices

For years, people uploaded photographs, videos and voice recordings believing they were sharing moments or building audiences. What many did not realise was that public material could become training data for artificial intelligence. Large AI systems may be developed using datasets containing personal information collected from websites and platforms. European regulators acknowledge that AI models can be trained on personal data, while the UK Information Commissioner says developers still need a lawful basis even when information comes from the open web.

Training a general model and cloning a person are not identical. The first learns patterns across datasets; the second uses targeted samples to reproduce an identifiable voice or appearance. Yet both raise the same question: who authorised the use?

A new proposition is emerging: instead of companies quietly collecting human material, people may knowingly license their faces, voices and movements for AI development or commercial digital replicas. In return, they could receive a one-time payment, royalties or a fee whenever their synthetic identity is used.

This is beginning. Shutterstock operates a contributor fund that compensates contributors when material from its library is licensed for AI model development. SAG-AFTRA has negotiated protections for digital replicas. The voice platform Narrativ allows performers to set prices and advertising preferences for AI-generated versions of their voices, with payment for approved uses.

At first glance, this is progress. People’s data has already been feeding digital systems, often without their knowledge, meaningful choice or financial benefit. A transparent marketplace could be fairer. At least individuals would know that a voice sample or facial scan was being collected. There would be a contract, and some value created from human identity could return to the human being.

It could also create work. A person might license a recorded voice for educational narration in several languages. A model could approve a digital replica for advertisements without attending repeated shoots. People far from production centres might gain opportunities previously unavailable to them.

But payment is not the same as protection. A person offered a modest amount may sign away far more than they understand. A broadly written agreement could allow a company to reproduce a face or voice for years, transfer it to other businesses, train additional models and place the replica in contexts the person would never endorse. The human receives one payment; the synthetic version may generate revenue indefinitely.

Faces and voices are not ordinary digital products. A password can be changed after a breach. A face cannot. A cloned voice can imitate a family member, support a fraudulent request or manufacture a statement that was never made. The US Federal Trade Commission has warned that scammers can create convincing voice clones from short audio clips found online. The answer is not to ban every voluntary AI licence. It is to establish non-negotiable human protections.

Consent must be specific, informed and renewable. Companies should state whether material will be used for training, identity replication or both. Agreements must define the product, audience, country, platform and time period. Permission for an educational video must not silently become permission for political advertising, gambling, medical claims, intimate material or religious messaging.

Compensation should continue when the replica continues earning. Where a digital identity is repeatedly used, the person should receive royalties or per-use payments, not merely a small initial fee. Individuals should be able to set minimum prices, reject industries and see where their replica appeared.

There must also be a genuine right to stop future use. Companies should delete source files, block new generations and notify third parties when a licence ends. Independent audits should test compliance. Strong security and penalties for leaks are essential because biometric material cannot simply be replaced.

Every synthetic performance should carry visible disclosure and a machine-readable marker. The European Union’s transparency rules, applicable from 2 August 2026, require certain AI-generated or manipulated material to be identifiable and deepfakes to be disclosed. Labels will not prevent every abuse, but they help preserve the distinction between human action and machine-generated imitation.

Children need stronger protection. Parents should not be able to permanently commercialise a child’s future identity. Any limited agreement should expire and require the individual’s fresh consent upon adulthood. Pakistan should treat this as urgent. As of 2026, the country still lacks a comprehensive enacted personal data protection law. This creates vulnerability just as international companies may seek diverse and comparatively inexpensive populations for AI training.  The greatest risk is not that humans will earn from AI. Fairly sharing value could correct years of silent extraction. The danger is that people will exchange permanent control for temporary payment.

As AI advances, societies must recognise the face, voice and digital personality as extensions of the human person, not raw material waiting to be harvested. Without consent, continuing compensation, traceability and the power to say no, we may lose more than jobs or privacy. We may lose the trust that allows us to believe what we see, hear and recognise. AI should expand human possibility, not gradually replace human ownership of the self.