AI identity verification is when a chatbot provider asks a user to prove who they really are — usually by uploading a government-issued ID, taking a live selfie, or both — before granting or restoring full access to the service. It’s different from just typing in a birthdate: the company checks a real document and, often, a facial geometry match against a specialized verification vendor, the same basic technique banks use to open an account.
How it actually works
Most AI companies don’t build this themselves. They plug in a third-party identity vendor — Persona is the one used by several major labs — that runs a know-your-customer-style flow: scan a passport, driver’s license, or national ID; take a live selfie or short video to prove a real person is present (not a photo of a photo); and generate a facial geometry template to match the ID photo to the selfie. That template counts as biometric data under several privacy laws, which is why companies are careful about how long they keep it and who else can see it.
This is a step up from “age prediction,” where a system just guesses whether an account likely belongs to a minor from writing style and usage patterns, without ever seeing an ID. Identity verification is the harder, more invasive version, usually reserved for specific situations rather than applied to everyone.
Who’s doing it, and when it kicks in
Anthropic is the first major consumer AI provider to formalize this. Starting July 8, 2026, Claude can require a government ID and a biometric face scan from Free, Pro, and Max users whose accounts have been flagged for a suspected policy violation — it’s offered as a way to prove your identity and keep your account rather than lose it outright. Team, Enterprise, and API customers aren’t affected. Anthropic says the data isn’t used to train its models and is handled by Persona rather than stored directly on Anthropic’s own systems.
OpenAI takes a lighter first step: ChatGPT tries to predict a user’s age from behavior and only asks an adult wrongly placed in the under-18 experience to verify with a selfie or ID to remove the restrictions. In most markets that’s optional; Italy is a notable exception, where OpenAI requires verification within a set window to keep full access. Character.AI, after lawsuits and state scrutiny over how its companion chatbots interacted with teenagers, runs a three-step check — behavioral signals first, a selfie second, and a government ID only as a last resort. Google’s Gemini, as of mid-2026, doesn’t require identity or age verification for standard consumer use.
Why this is happening now
The push has two sources. One is legal exposure: AI companion apps have faced wrongful-death lawsuits and state investigations tied to interactions with minors, and regulators want proof that providers are actually keeping children off adult-oriented or unsupervised chatbot features. The other is pending legislation: the GUARD Act, a bipartisan bill that unanimously cleared the US Senate Judiciary Committee, would require any company offering AI companion chatbots to the public to verify every user’s age and block minors outright, plus force chatbots to disclose they aren’t human. It hasn’t passed the full Senate or House yet, so nothing in it is law today — but it signals where consumer-facing AI products may be headed regardless of what any single company chooses to do voluntarily.
The privacy trade-off
Collecting a government ID and a face scan solves one problem (proving who someone is) by creating another (a valuable, hard-to-change piece of data sitting somewhere). Facial geometry is treated as sensitive biometric data in a number of jurisdictions — Illinois’s Biometric Information Privacy Act, for instance, allows penalties of $1,000 to $5,000 per violation for collecting it without proper consent — which is a large part of why companies route verification through specialized vendors rather than building it in-house, and why they emphasize what they don’t keep. Critics of mandatory approaches like the GUARD Act also raise a separate concern: applying ID checks to every user, not just flagged accounts, means adults who want to use a chatbot anonymously no longer can.
What to do if you’re asked to verify
Currently, most people will never see this prompt — it’s triggered by a flagged account, a specific under-18 correction, or a market-specific law, not by default. If you are asked, verify only through the official in-app or in-site flow (never a link sent by email or text), expect the process to reject photocopies, screenshots, or digital ID scans, and check the provider’s help pages for exactly what’s collected and how long it’s kept — see Claude’s own explanation of its verification process as an example of the level of detail to look for.
In the news
For the specifics of the first large-scale rollout, see our report on Anthropic’s decision to require government ID and biometrics from some Claude users.
FAQ
Do I have to verify my identity to use ChatGPT or Claude? No, not by default. Verification currently applies only to flagged accounts, users incorrectly placed in an under-18 experience, or specific markets with their own rules — not to the general user base.
Is this the same as age verification? Related but not identical. Age verification (or age prediction) tries to estimate whether you’re over or under 18, sometimes without any document at all. Identity verification confirms exactly who you are using a government ID and biometric match, which is a higher bar.
What happens to my ID photo and face scan afterward? It depends on the company and vendor. Anthropic, for example, says verification data is processed by Persona and isn’t used to train its models; specifics on retention and deletion are typically listed in the provider’s privacy policy or help center.
Could ID verification become mandatory for everyone? Possibly, if legislation like the GUARD Act becomes law — but as of mid-2026 it’s still moving through the US Senate and hasn’t been enacted, so no AI chatbot is currently required by federal law to verify every user.
Sources: Anthropic’s Claude identity verification help page, TechCrunch on Anthropic’s ID policy, OpenAI on age prediction in ChatGPT, GUARD Act text, Congress.gov.