Anthropic's Claude Now Asks for Government ID in Some Cases, Raising Privacy Concerns

Anthropic, the AI company behind the Claude chatbot, updated its privacy policy in mid June to require identity verification from a small subset of users. Starting July 8, users whose accounts are flagged for potential fraud may need to upload a government issued passport or driver's license to appeal. The policy, first spotted by TechCrunch, also allows Anthropic to collect a selfie photo and generate a face geometry template for biometric matching.

The company says the change applies to users whose accounts are flagged but not yet banned. Anthropic has tens of millions of monthly users, though it declined to say how many would be affected. The policy shift positions Anthropic at the center of ongoing tensions between AI companies and the US government over model security.

Verification process and data retention

Anthropic is using San Francisco based Persona as its identity checking provider. The policy says Persona collects a photo scan of a government ID, a selfie photo or video, and a digitized face geometry template. That template is considered biometric data under Illinois law and may be protected in other states with similar privacy rules.

• Anthropic says it will keep a record of the verification result, including whether the user has reached a certain age.
• The company says it decides how long Persona retains identity documents, but did not disclose specific deletion timelines.
• For comparison, Roblox uses Persona and deletes user images immediately after processing, limiting the risk of leaks.
• Persona is backed by Founders Fund, the investment firm founded by Trump supporter Peter Thiel, who also invests in Anthropic.
• Discord chose Persona for age verification earlier this year, then reversed the decision after user backlash.

Broader government feud and security concerns

The verification policy arrives as Anthropic remains locked in a standoff with the Trump administration. In April, the company released Mythos, an AI model it says can find security vulnerabilities. Reports claim Mythos cracked NSA classified systems in hours. The Department of Defense later designated Anthropic a supply chain risk, apparently in retaliation for the company refusing to allow its technology to power mass surveillance or fully autonomous weapons.

Trump officials effectively forced Anthropic to pull its cybersecurity models over concerns that a jailbreak could break their guardrails. The White House has accused Anthropic of not cooperating on AI safety measures. Anthropic, for its part, says it is complying with legal requirements and that the identity checks are part of a routine appeals process for flagged accounts.

The policy change is scheduled to take effect on July 8. It is unclear how many users will encounter the verification prompt or whether the company will extend the checks to more use cases in the future. Critics note that Persona remains subject to US government demands for stored user data, adding another layer of privacy risk for users who comply with the request.