Protecting Your Likeness in 2026: A Complete Guide

In 2026, your likeness is a digital asset, one that AI systems can extract, replicate, and deploy at scale, with or without your consent. This guide covers what protection actually means, what the law currently covers, and the practical steps to take right now.

What "likeness" means legally

Your likeness includes:

• Face, your distinctive facial features, recognizable appearance
• Voice, your vocal signature, speech patterns, timbre
• Name, your real name or professional name/brand
• Signature style, in some jurisdictions, a recognizable visual aesthetic or pose

The legal protection for each varies by jurisdiction. In the US, Right of Publicity laws protect your likeness from commercial use without consent, but only in about 35 states, with widely different scope. Federally, the proposed NO FAKES Act would create a national standard for AI replicas specifically.

The three-layer protection model

Effective likeness protection works in three layers, each covering what the others miss:

Layer 1: Registration and provenance. Before any violation occurs, establish an immutable record that you own your likeness. This means:

• Registering your biometrics (face embedding, voice fingerprint) with a timestamped service
• Applying C2PA content credentials to your reference photos, cryptographic certificates backed by Adobe, Microsoft, and Google
• Optionally, copyright-registering your professional photos ($55 per batch via the US Copyright Office)

Layer 2: Continuous monitoring. Violations you don't know about can't be removed. You need automated surveillance across:

• 50+ social platforms (Google Images, Instagram, TikTok, Reddit, YouTube, adult sites, etc.)
• Telegram channels and private groups (requires MTProto API access, the Bot API doesn't reach private channels)
• The Dark Web (Tor-accessible forums and markets where deepfakes are sold)

Layer 3: Automated enforcement. Every violation that isn't removed within hours spreads exponentially. Manual DMCA filing takes 30–90 days per violation. Automated enforcement, filing the moment detection happens, compresses this to hours.

AI opt-outs: which ones matter

Several AI training opt-out mechanisms exist. In rough order of impact:

1. Spawning's "Do Not Train" registry, used by Stable Diffusion, Midjourney, and others. Apply at spawning.ai.
2. Adobe's Content Authenticity Initiative, opt-out for Adobe Firefly training
3. Getty Images opt-out, if your photos are licensed through Getty
4. LAION removal request, LAION-5B is the dataset most open models were trained on; request removal at laion.ai

None of these work retroactively on models already trained on your data. For retroactive compensation, the legal theory is unjust enrichment, companies profited from your likeness without consent. Several class actions are currently working through the courts on this theory.

New laws that matter in 2026

EU AI Act Article 50 (enforced from August 2026): All synthetic media of real people must be labeled as AI-generated. Violations carry fines up to €15M.

ELVIS Act (Tennessee): First US law specifically targeting AI voice cloning. Applies to voice content distributed in Tennessee; damages plus punitive damages.

California AB 2602: Film and TV studios must obtain separate written agreements before using AI replicas of performers.

Australia Online Safety Act / NCII Scheme: eSafety Commissioner can compel platform removal within 24 hours for intimate deepfakes. AUD $700,000/day fines for non-compliant platforms.

What to do right now

In order of impact:

1. Upload a reference photo and voice sample to register your biometrics
2. Enable C2PA signing on your reference photo
3. Set up monitoring across the platforms where violations are most likely (for most creators: Instagram, TikTok, Reddit, Telegram)
4. File opt-outs with the major AI training registries
5. Document your current online presence, screenshots, cached pages, to establish a baseline for future damages claims

Protection is most effective when it starts before a violation, not after. The evidence you have from before an incident is worth far more legally than what you can reconstruct after the fact.