How to Protect Your Face and Voice From AI

Seven practical steps to protect your face and voice from AI misuse, from a free evidence folder to takedown routes and the laws that apply in 2026.

Royall publishes this guide. Descriptions of other companies come from their own public pages, accessed October 4, 2026.

Quick answer

You can't make your face or voice impossible to copy, but you can make misuse easier to prove, find and remove. Keep dated originals in a secure folder, write down which AI uses you would allow, watch the places where you're most visible, save evidence before you report anything, and use the removal route that fits the harm. US law now gives you a 48-hour takedown right for intimate deepfakes, and several states cover AI copies of your voice and likeness.

If a fake of you is up right now

Save evidence first. Copy the URL, account name, date and time, and take screenshots or a screen recording.

Report it to the platform using its impersonation, privacy or altered-media form.

Intimate image, real or AI-made? Covered platforms must remove it within 48 hours of a valid request under the TAKE IT DOWN Act. If a platform fails to act, report it to the FTC at takeitdown.ftc.gov. Our TAKE IT DOWN Act explainer explains what a valid request needs.

If you were under 18 in the image, use NCMEC's Take It Down.

If you're an adult, StopNCII.org creates a hash (digital fingerprint) on your device so participating platforms can detect and block the image.

If you're threatened or in danger, contact local police (911 in the US).

Questions for Royall: royall.ai/help or support@royall.ai.

Step 1: How do I prove what my real face and voice look like?

Keep high-quality originals of your face and voice with reliable creation dates: raw phone video, voice memos, photos straight off the camera. Back them up in two places. The point is being able to show what is genuinely yours, and when it existed.

The no-upload version: your protection folder. This is the full version of the folder other Royall guides refer to.

Create an encrypted folder (most phones and computers can do this) with a second backup.
Add originals: a few minutes of natural speech and short face videos from several angles, plus any professional footage you own.
Add proof of ownership and identity: contracts showing you own or control footage, and a copy of your ID if a representative may need to act for you.
Add your consent rules (Step 2) as a dated document.
Add licenses and releases you have signed.
Add an incident log for anything suspicious (Step 4).
Tell a trusted person or your lawyer where it is and how to open it.

The tradeoff: an offline folder can't watch the web for you or answer an app that asks for permission.

Step 2: Which AI uses of my likeness would I actually allow?

Decide before anyone asks. Would you allow your face or voice in ads, dubbing, games, audiobooks, AI model training, or none of these? Treat training (teaching a model your likeness) and output (generating new media of you) as separate permissions. Write your answer down, date it, and update it when your view changes. For real deals, use a specific written agreement, not a statement on your website.

Step 3: Where should I monitor for fakes?

Combine methods. No scanner sees the whole internet.

Platform tools. YouTube's likeness detection is free for eligible creators over 18. It is experimental, matches faces only for now (YouTube says it aims to add audio), scans newly uploaded videos, and requires a government ID and a selfie video. YouTube Help
Monitoring services. Several offer free plans. We list them, with prices, in our alternatives guide.
People who know you. Friends and followers often spot impersonations first. Give them one place to send links.
Manual checks. Search your name, usernames, common misspellings and catchphrases every week or month.

Step 4: What should I save before I report a deepfake?

Before it disappears, save the URL, the account name, the date and time, screenshots or a screen recording, any messages or ads involved, the file itself where that is lawful and safe, and a note on why you know it isn't you. Log each incident in one place. Our rights management guide has a ready-made incident log template.

Step 5: Which removal route should I use?

Situation	Route
Someone is pretending to be you	The platform's impersonation or privacy form
An intimate image of you, real or AI-made	The platform's TAKE IT DOWN request process; then takeitdown.ftc.gov if it isn't removed
Your own photo, video or recording was copied	A copyright notice, if you own or control the copyright
Your face or voice is selling something	Right of publicity or false endorsement law, which varies by state; talk to a lawyer

Copyright protects recordings you own. It does not protect your face or voice as such. Our takedown guide covers each route in detail.

Step 6: Should I add Content Credentials to my own posts?

It can help. C2PA Content Credentials attach tamper-evident information about how a participating file was made and edited, so people can check your originals. They don't prove consent, and a missing credential doesn't mean a file is fake. Treat provenance as one trust signal. C2PA specifications

Step 7: How do I keep track of who has permission?

Keep a register of every license: who, what use, how long, where, and whether training is included. A spreadsheet works; our rights management guide has the template. If you license often, a consent platform can let apps check your current choices automatically.

What laws protect my face and voice from AI?

Status as of October 4, 2026:

TAKE IT DOWN Act (federal, Public Law 119-12). Covered platforms must remove an intimate image published without consent, including an AI "digital forgery", within 48 hours of a valid request and make reasonable efforts to remove known identical copies. The FTC has enforced the platform duties since May 19, 2026. Law text · Tracker
Tennessee ELVIS Act (Pub. Ch. 588, in effect July 1, 2024). Protects your voice "regardless of whether the sound contains the actual voice or a simulation of the voice", which brings AI clones within the law. News, commentary, satire and parody are treated as fair use to the extent the First Amendment protects them. Pub. Ch. 588 · Tracker
California SB 1111. Signed September 30, 2026 (Chapter 862, Statutes of 2026), effective January 1, 2027. It states that California's right of publicity statute (Civil Code §3344) covers digital replicas, and that using a replica to impersonate someone counts as false impersonation. Bill history · Tracker
NO FAKES Act (federal, proposed). Not law. The Senate Judiciary Committee approved S. 4591 by voice vote on June 18, 2026, and it was reported on June 24 (Calendar No. 446) with no floor vote yet. The House bill, H.R. 8915, is in committee. S. 4591 on congress.gov · Tracker

Rules differ by state and country. Browse the Identity Rights Tracker for where you live.

Where Royall fits

Royall is one way to handle Steps 1, 3 and 7 in a single account. You record your face and voice in the browser, sign one consent, and choose which uses you allow. Sentinel checks reported links against your recordings and sends takedown notices for you. Apps that integrate Royall can ask before using your likeness. Your recordings aren't public, but they are not only for you: Royall's Terms let it use your likeness to train its own detection models, and licensing is offered separately with your consent. Here is what the Terms and Privacy Policy say, so you can decide before signing up.

How Royall handles your data What Royall's Terms of Service and Privacy Policy say today (both effective September 15, 2026):

Royall's own models. Royall may use your likeness to train Royall-operated AI models to improve its services, for example its likeness detection (Terms §1.4).

Licensing to partners. Royall may offer you opportunities to license your recordings to Royall partners, including for AI training, under separate terms and with your consent (Terms §1.2 and §1.5; Privacy Policy). Those licenses must bar the partner from using your likeness to train models for deepfake generation, sexual content, harassment, infringement or any illegal purpose (Terms §1.5).

Revoking consent stops future licenses. It cannot pull your likeness out of data sets already shared with third parties under earlier licenses (Terms §1.3; Privacy Policy).

Retention. Biometric data is kept as long as needed to provide the services you request, or up to three years after your last interaction, unless you revoke consent (Privacy Policy).

Access and deletion. Ask at privacy@royall.ai (Privacy Policy).

Apps that check permission through Herald or Connect get a yes or no answer. Royall says partners using Connect do not receive your recordings (royall.ai).

Key takeaways

Keep dated originals and a written list of what you allow.
Save evidence before you report.
Match the removal route to the harm.
Intimate deepfakes have a 48-hour federal takedown route.
You don't need to upload your likeness to a company to take real protective steps.

FAQ

Can I stop AI from cloning my voice? Not completely. You can make clones easier to find, prove and challenge, and in Tennessee the ELVIS Act covers simulated voices.

Should I upload my face and voice to a protection service? It's a personal risk decision. Read the provider's terms on training, licensing and retention first, using our vendor checklist. A private evidence folder is a valid alternative.

Is every AI use of my likeness illegal? No. It depends on where you are, the purpose and any contract. News, commentary, parody and endorsements are treated differently.

What if the fake is of my child? For intimate images of anyone under 18, use NCMEC's Take It Down and report to the platform and police.

Sources

Related guides

This is general information, not legal advice.