Anthropic Starts Requiring Govt IDs and Video Selfies From Claude Users

Anthropic has introduced identity verification requirements for certain users of its Claude platform, marking a shift in how access to some features is managed.

The company is asking selected users to provide a government-issued photo ID along with a live selfie when accessing specific capabilities. According to Anthropic, this is part of "platform integrity checks" and broader safety and compliance measures.

The policy is not applied universally. Verification is triggered only in certain cases, including when activity appears potentially fraudulent or abusive.

Anthropic is using Persona as its verification partner. Users must submit a valid passport, driver's license, or national ID card. Photocopies, mobile IDs, and student IDs are not accepted.

The company said that identity data is processed through Persona's systems rather than stored directly by Anthropic. The data is encrypted during transfer and storage, excluded from model training, and not used for marketing purposes.

The move has drawn criticism from users, particularly those who had previously viewed Anthropic as a privacy-focused alternative. The requirement is not tied to any government mandate, but rather a decision made internally by the company.

Concerns have also been raised about the risks of storing sensitive identity data with third-party providers. Past incidents, including breaches involving similar verification systems, have highlighted potential vulnerabilities.

Anthropic saw a surge in users earlier this year, including users switching from OpenAI. At the time, the company reported strong growth in signups and increased usage of its free tier.

The introduction of verification measures comes as Anthropic continues to expand safety controls across its platform.

The verification requirement aligns with other measures introduced by Anthropic, including systems to detect underage users and enforce regional access restrictions.

These controls have led to some account suspensions, including cases where users reported being incorrectly flagged and losing access to their data.