Summary: TechCrunch reported on June 22, 2026 that Anthropic’s privacy policy could require flagged Claude users to submit government ID, with new wording effective from July 8. The process may include passport or license scans, selfies or video, and face geometry data handled through Persona. The article says the uncertainty around triggers, retention, and government access makes this a compliance and privacy risk, not just a product feature.
Writing on June 22, 2026, TechCrunch security editor Zack Whittaker flagged a quiet revision to Anthropic’s privacy policy: “in certain circumstances,” the company can now ask Claude users to hand over a government-issued document proving their age or identity. The revised wording went into the policy on June 17 and applies from July 8. Once the flow triggers, the user must provide scans of a passport or driver’s license along with a selfie photo or video, and those images get turned into a digitized face geometry template. With tens of millions of people on the service each month, the AI company has effectively imported bank-style identity checks into a consumer chatbot.
What Changed in the Terms
The policy names a wide set of purposes: administering accounts, holding users to the terms of service, fighting fraud, abuse, and unlawful conduct, and looking into security issues. Anthropic spokesperson Michael Aciman cited a post by company lead Thariq Shihipar, explaining that the policy “was updated on June 17 as an update to the appeals process” and is “unrelated to the Fable or Mythos rollout.” The people who will actually meet this process are a “small subset of users” whose accounts were flagged without rising to a ban — accounts that used to be suspended outright now get another route: confirm who you are and keep access. How many people that subset covers, and what situations the “certain circumstances” describe, Anthropic would not say; it noted only that users might “see a verification prompt when accessing certain capabilities,” framing the whole thing as routine platform integrity work. That vagueness about triggers is exactly what unsettles privacy advocates.
Face Geometry Data: Retention Questions and Access Rights
Where the data ends up is the real fight. License scans, passport scans, selfies, and face geometry templates all count as biometric data, and jurisdictions such as Illinois give face geometry information its own legal safeguards. Anthropic says it holds onto a record of the verification result — for example, whether a user is past the required age. The checks themselves run through Persona, a San Francisco company, and how long data stays on Persona’s side is Anthropic’s call — yet the company would not say when deletion happens. The contrast is Roblox, another Persona customer, which pledged in its age-check announcement that user images are deleted “immediately” after processing. And one reality cannot be dodged: whatever Persona stores for U.S. users stays exposed to U.S. government demands for stored user information. No defined retention window, a third party holding the files, and government access — put those together and you have the textbook risk profile for biometric data.
Persona, the Outside Vendor Running the Checks
Persona, the startup doing the verification, is backed by Founders Fund — and Peter Thiel invests in Anthropic as well, a connection that has already drawn criticism. The market track record is bumpy too: in February, Discord said it would use Persona to age-verify users worldwide, then pushed back the global rollout within weeks under heavy user protest. Consumer platforms adopting identity checks is clearly the direction the industry is heading, yet nearly every rollout arrives with a privacy fight attached — what separates vendors is how openly they talk about data retention.
Age Checks Against a Politicized Backdrop
Claude’s floor is age 18, and Anthropic stood up age-assurance checks earlier this year to satisfy mandates from several states and countries — the most plausible regulatory context for this update. The timing still invites political readings: on June 13, the White House made Anthropic pull its latest cybersecurity models over alleged jailbreak concerns, and in early March the Pentagon branded Anthropic a “supply chain risk.” In a climate that tense, any policy change that widens the collection of users’ biometric data will be put under a microscope, however many times the company repeats that it belongs to the appeals process.
The Practical Takeaways
Three consequences are worth weighing. First, anyone whose account is flagged must pick between verifying and losing it, and the smarter move is to learn where the data travels and for how long before handing over an ID. Second, enterprise buyers should read this as a warning shot: consumer AI is converging on finance-style identity checks, so biometric retention policies, third-party processors, and data jurisdiction all belong on the vendor review checklist. Third, statutes like Illinois’s, together with age-assurance laws around the world, will keep marking out where these mechanisms may legally operate. Identity checks are no longer just a product feature; they are now a matter of contracts, legal review, and compliance.