You Switched to Claude Over Surveillance Fears. Now It Wants Your Passport

Anthropic softly published identity verification requirements for Claude this week, asking definite users to manus implicit a government-issued photograph ID and a unrecorded selfie. Something its competitors don’t require.

“We are rolling retired individuality verification for a fewer usage cases, and you mightiness spot a verification punctual erstwhile accessing definite capabilities, arsenic portion of our regular level integrity checks, oregon different information and compliance measures,” Anthropic said. “We lone usage your verification information to corroborate who you are and not for immoderate different purposes.”

Millions of users fled OpenAI for Anthropic in February after OpenAI signed a woody to deploy AI connected Pentagon classified networks—a declaration Anthropic turned down implicit concerns astir wide surveillance and autonomous weapons. Daily signups broke records, and escaped users were up 60% since January, Anthropic said astatine the time. The privacy-conscious assemblage had recovered its home.

That crowd, it seems, whitethorn present person immoderate documents to hole if it wants to proceed utilizing Claude. The reactions truthful acold person been rather negative, pointing retired that it’s a deliberate determination and not a regularisation oregon a mandatory bid imposed by a authorities connected Anthropic arsenic a work provider.

According to the assistance halfway page, which went unrecorded connected April 14, Anthropic selected Persona Identities arsenic its verification partner—the aforesaid KYC infrastructure utilized crossed fiscal services—and requires a physical, undamaged passport, driver's license, oregon nationalist individuality card. Photocopies, mobile IDs, and pupil credentials don't count. A unrecorded selfie whitethorn besides beryllium required.

The argumentation isn't cosmopolitan yet. Verification volition trigger erstwhile accessing "certain capabilities," during "routine level integrity checks," oregon arsenic portion of information and compliance measures. Anthropic hasn't said publically which features are gated, oregon what idiosyncratic behaviour mightiness punctual a check. The institution did not instantly respond to Decrypt's petition for further details.

On information handling, Anthropic draws a cautious line: your ID and selfie spell to Persona's servers, not Anthropic's ain systems. The institution says it is the information controller mounting the terms, and that Persona tin usage the accusation to verify individuality and amended fraud detection. The information is encrypted successful transit and astatine rest, excluded from exemplary training, and won't beryllium shared with 3rd parties for marketing, thing Anthropic has been careful to promise since its earliest commercialized policies.

Careful promises, though, person a past of gathering careless infrastructure. An October 2025 breach astatine Discord exposed astir 70,000 authorities IDs users had submitted for property verification. Persona is simply a superior subordinate successful this space, but third-party custody of authorities documents has demonstrated repeatedly that nary 3rd enactment is immune.

Tighter individuality controls besides acceptable a signifier Anthropic has been gathering toward. In December, the institution announced classifiers to observe users who self-identify arsenic minors. Multiple big users had their accounts suspended anyway, reporting that full task histories were wiped portion they tried to entreaty incorrect flags.

Accounts registered from regions Anthropic doesn't formally service are besides taxable to bans—a item that lands hardest connected Chinese users accessing Claude done intermediaries, since a unrecorded selfie matched against a carnal authorities papers is hard to fake your mode through.