Anthropic may ask some Claude users to verify their age or identity by uploading government-issued documents, under a new privacy policy that takes effect on July 8. The policy adds a “Verification Data” category covering an image of a passport, driver’s license or national ID, a selfie photo or video, and a facial geometry template that Anthropic acknowledges “may be considered ‘biometric data’ in some jurisdictions.”
The requirement applies only to consumer accounts on the Free, Pro and Max tiers, not to Team, Enterprise or API customers. The company has tens of millions of monthly users and says only a small fraction will be affected.
Anthropic describes the checks as an appeals mechanism rather than a new barrier. A spokesperson said the policy gives users whose accounts are flagged for potential fraud or policy violations, but not outright banned, a way to appeal by proving who they are instead of losing access.
The company said the June 17 update to its identity-verification policy was “unrelated to the Fable or Mythos rollout,” referring to the models it pulled this month. Anthropic uses the firm Persona for full identity checks and a separate vendor, Yoti, for age-only checks that return a simple pass or fail. It says verification data is encrypted, is not used to train its models, and that Claude remains ad-free with no sale of user data.
The change has drawn scrutiny on several fronts. Persona, which processes and stores the documents on its own servers, is backed by Founders Fund, the firm led by Peter Thiel, who is also an Anthropic investor.
Discord picked Persona for age checks earlier this year, then reversed course after user backlash over the Thiel link. Privacy advocates note that facial geometry is sensitive, that Persona can still face US government data demands, and that Anthropic has not specified how long the documents are retained, unlike some Persona customers that delete images immediately. Anthropic is the first major AI maker to formally codify biometric identity collection at the consumer level, while OpenAI and Google do not require it for standard access.
Why It’s Drawing Scrutiny
For a company that markets itself on safety and responsible AI, asking users to hand passport scans and facial biometrics to a Thiel-tied vendor raises a trust question that technical assurances alone may not settle.
The opacity compounds it: the policy does not say what triggers a check, what happens if a user refuses, or when data is deleted. Some users have already reported being wrongly flagged as minors and suspended, then asked to verify with ID.
Legal exposure is real too, since states like Illinois treat facial geometry as protected biometric data with per-violation penalties, and non-US transfers raise unresolved questions under laws such as India’s data-protection act.
The Backdrop With Washington
The timing is hard to separate from Anthropic’s standoff with the Trump administration. This month, US officials effectively forced the company to pull its Fable 5 and Mythos 5 models over a claimed jailbreak, and in March the Pentagon labeled Anthropic a supply-chain risk, after it declined to allow its technology for mass domestic surveillance or autonomous weapons.
Some analysts speculate that verifying users could let Anthropic restore restricted models for US persons only, satisfying export concerns, though the company denies a link and the Commerce Department has not commented. President Trump signaled last week, after a G7 meeting with CEO Dario Amodei, that he no longer views Anthropic as a national security threat, leaving the broader pressure unresolved.
