Claude bids farewell to the anonymous era: Anthropic will introduce real-name facial recognition, allowing data sharing with police without subpoenas

BlockBeats News, June 15 — Anthropic recently sent a privacy policy update notification to users, announcing the implementation of a new Terms of Service starting July 8, 2026. The new policy will not only introduce real-name and facial recognition authentication but also lower the threshold for disclosing user data to law enforcement agencies, marking a further end to the era of anonymity in large model services.

According to the new regulations, Anthropic may require individual users of Claude Free, Pro, and Max to complete identity verification in specific scenarios such as detecting account abuse, violations, or suspected underage use. The verification process is conducted through third-party compliance service provider Persona, where users must upload government-issued ID documents and take a real-time selfie for comparison. Anthropic emphasizes that biometric and document data uploaded by users will not be stored on Anthropic’s own servers and will not be used for model training or marketing.

A more critical change is the adjustment of data sharing clauses. The old policy required disclosure of user data only when "legally mandated." The new regulation states that as long as Anthropic has a "good faith belief" that disclosure is "reasonably necessary" to prevent harm to persons, fraud, illegal activities, or to uphold platform terms, it can proactively share user conversations and data with law enforcement or government agencies. The data sharing clause adjustment applies only to personal consumer accounts; data for enterprise and API clients remains protected under commercial confidentiality agreements.

As early as June 9, Anthropic had enforced a 30-day mandatory data retention policy for the most powerful models such as Fable 5 and Mythos 5, to investigate jailbreaking behaviors. While user verification and proactive sharing clauses have long been standard for tech giants like OpenAI and Google, the recent security tightening measures—especially under the pressure of export restrictions on the most powerful models and the banning of Fable 5 accounts—continue to prompt sensitive users and enterprises to reassess Claude’s privacy boundaries.