Governance document

CA-D-000012

Anthropic Privacy Policy (Superseded Capture)

Anthropic

Last change detected July 9, 2026 Privacy policy

SHA-256: e91b78d120f18b8a635… 4 versions captured 2 changes documented

Get the research letter

Companies change their terms quietly. We read every version and catch what actually changed. One email a week on the changes that matter and what they mean.

View original document at Anthropic ↗

This page describes what the document states, permits, or reserves. It does not constitute a legal determination about enforceability. Regulatory applicability may vary by jurisdiction. Methodology

81 Total

14 High severity

50 Medium severity

17 Low severity

Key Facts

What may result in actions with effects outside the Services depending on the permissions a user grants Claude?

Depending on the permissions a user grants Claude, some Outputs may result in actions with effects outside the Services, such as sending communications, modifying files, or interacting with third-party services on the user's behalf.

Can Outputs result in sending communications, modifying files, or interacting with third-party services on the user's behalf?

Does Anthropic control the data practices of Third-Party Services?

Anthropic does not control, and is not responsible for, the data practices of Third-Party Services.

Is Anthropic responsible for the data practices of Third-Party Services?

Anthropic does not control, and is not responsible for, the data practices of Third-Party Services.

What does Anthropic collect for identity verification purposes?

Anthropic collects biometric data for identity verification purposes, including a government-issued identity document image, a photo or video image of the user, and facial geometry templates, which may be considered biometric data in some jurisdictions.

What biometric data does Anthropic collect?

What will Anthropic do if it investigates reports of personal data provided by children under 18?

Anthropic will investigate reports of personal data provided by children under 18 and, if appropriate, delete that personal data.

Does Anthropic have measures in place to detect and remove children from its Services?

Anthropic has measures in place to detect and remove children from its Services.

With whom may Anthropic share personal data where it has a good-faith belief that disclosure is reasonably necessary to comply with applicable law?

Anthropic may share personal data with government authorities, law enforcement, or other third parties where it has a good-faith belief that disclosure is reasonably necessary to comply with applicable law.

May Anthropic use a user's Inputs and Outputs to train Anthropic AI models?

Anthropic may use a user's Inputs and Outputs to train and improve Anthropic AI models, unless the user opts out through their account settings.

Stay ahead of the changes

Track Anthropic and get the diff the day its terms change.

Monitor Anthropic + up to 20 platforms — $4.99/mo →

Summary

This document explains how Anthropic collects and uses your data, including the content of your conversations with Claude. By default, Anthropic uses your chats to train its AI—you can turn this off in your account settings, though flagged or self-reported content can still be used even if you opt out. Anthropic may also collect biometric data like your photo and facial geometry for identity verification, and may share your data with law enforcement or send it to third-party services when Claude acts on your behalf.

Analysis

This privacy policy establishes Anthropic's data collection, usage, sharing, and user-rights framework for its Services. Anthropic collects biometric data including government-issued ID images, photos or videos, and facial geometry templates for identity verification. By default, Anthropic may use users' Inputs and Outputs to train and improve its AI models, subject to an opt-out available through account settings; however, that opt-out does not apply to content flagged for safety review or content users explicitly report to Anthropic. Anthropic may share personal data with government authorities or law enforcement based on a good-faith belief that disclosure is reasonably necessary for legal compliance, and may transmit Inputs, Outputs, and instructions to Third-Party Services to perform agentic actions on users' behalf, for whose data practices Anthropic disclaims responsibility. Anthropic does not engage in consequential decision-making based solely on automated processing or profiling.

What this means for you

As an individual user, your conversation Inputs and Outputs are used by Anthropic to train its AI models by default. You can opt out of this by changing your account settings, though content flagged for safety review or content you explicitly report to Anthropic remains available for model improvement regardless of that opt-out. Anthropic collects sensitive biometric data—including your photo, government ID image, and facial geometry templates—for identity verification. When you grant Claude permissions to act on your behalf, your Inputs, Outputs, and instructions may be sent to Third-Party Services outside Anthropic's control, and Anthropic is not responsible for how those services handle your data. Anthropic may also disclose your personal data to government authorities or law enforcement based on its own good-faith assessment of legal necessity, without your consent. One concrete action available to you: navigate to your account settings to opt out of having your Inputs and Outputs used for model training.

Institutional Analysis

Stay ahead of the changes

Institutional analysis available with Insight

Which mapped governance frameworks each document engages, tied to the specific provisions that engage them.

Monitor Anthropic + up to 20 platforms — $4.99/mo → Get the research letter, free

2 important changes detected

4 versions captured · Last updated: July 2026

July 9, 2026

low

What changed Anthropic's privacy policy was updated on July 9, 2026 to add a single introductory sentence describing the company as 'an AI safety and research company working to build reliable, interpretable, and steerable AI systems.' This sentence was removed from its previous location and then reinserted in the document header. The change is editorial and does not alter any substantive privacy rights, data practices, or user obligations.

Why this matters This change has no material impact on consumer privacy rights, data collection practices, or obligations. The updated policy now includes a single introductory sentence identifying Anthropic as an AI safety and research company. This is a purely editorial modification with no changes to how personal data is handled, retained, or shared.

View full change record →

June 9, 2026

low

What changed Anthropic updated its Privacy Policy on June 9, 2026, with an effective date of July 8, 2026. The revised policy narrows its scope to clarify that it applies when Anthropic acts as a data controller for consumer services (Claude.ai, Claude Team), but not when it processes data on behalf of enterprise customers whose own data governance controls how their data is handled. The policy also expands the definition of 'Inputs' to explicitly include content submitted through third-party integrations and connected services, and clarifies that personal data or external references in Inputs may be reproduced in Outputs.

Why this matters The updated policy clarifies that it applies when you use Anthropic's consumer services like Claude.ai or Claude Team, but does not govern your data when you access Anthropic products through your employer or an enterprise account. In that case, your employer's or enterprise customer's own data governance policies control how your data is handled. The policy also now explicitly states that personal data or external references included in your inputs to the service may be reproduced in the outputs generated in response.

View full change record →

Featured, High severity

Agentic outputs may affect third-party services Compare →

Claude's agentic outputs can have tangible real-world consequences beyond the Anthropic platform, and those consequences are conditioned on permissions the user grants, placing meaningful weight on the user's permission decisions.

Added July 10, 2026 Unique
Anthropic not responsible for third-party data practices Compare →

This disclaimer means users have no recourse against Anthropic for how Third-Party Services collect, use, or share data, even when those services are accessed through or in connection with Anthropic's platform.

Added July 10, 2026 Standard Seen across 287 platforms
Biometric data collection for verification Compare →

Facial geometry templates are among the most sensitive categories of personal data and are regulated as biometric data in several jurisdictions; their collection by Anthropic carries significant privacy implications depending on applicable law.

Added July 10, 2026 Standard Seen across 296 platforms
Deletion of child data upon notification Compare →

Deletion is conditional on Anthropic's determination that it is 'appropriate,' meaning it is not automatic upon notification; the outcome depends on Anthropic's investigation.

Added July 10, 2026 Standard Seen across 277 platforms
Detection and removal of underage users Compare →

Anthropic maintains measures to detect and remove children from its Services.

Added July 10, 2026 Standard Seen across 277 platforms
Disclosure to government and law enforcement Compare →

This provision permits disclosure of personal data to government and law enforcement actors without user consent, conditioned only on Anthropic's good-faith belief that it is reasonably necessary for legal compliance.

Added July 10, 2026 Standard Seen across 288 platforms
Inputs and outputs sent to third-party services Compare →

User data—including the content of interactions with Claude—may be transmitted to external third parties outside Anthropic's direct control, expanding the set of entities that receive and process that data.

Added July 10, 2026 Standard Seen across 288 platforms
Inputs and Outputs used for model training Compare →

Training use of Inputs and Outputs is the default; users must take an affirmative step through account settings to prevent it, meaning inaction results in data use for model improvement.

Added July 10, 2026 Standard Seen across 279 platforms
Safety-flagged content training override Compare →

The opt-out for model training is not absolute; two specific conditions—safety flagging and user-initiated reporting—override the opt-out, meaning users cannot prevent training use of flagged or self-reported content.

Added July 10, 2026 Standard Seen across 279 platforms
User opt-out for model training Compare →

The availability of an opt-out mechanism gives users a degree of control over whether their interaction data is used for model training, though that control must be exercised affirmatively through account settings.

Added July 10, 2026 Standard Seen across 290 platforms

Featured, Medium severity

Flagged content disassociated but re-identifiable for enforcement Compare →

The disassociation provides only conditional anonymity: Anthropic preserves the operational capacity to re-identify flagged content, meaning the separation is not permanent and can be reversed for enforcement purposes.

Added July 10, 2026 Standard Seen across 279 platforms
No automated legal decision-making Compare →

This commitment limits Anthropic's use of fully automated decision-making for consequential outcomes, which is directly relevant to rights under privacy frameworks such as GDPR that restrict such practices.

Added July 10, 2026 Standard Seen across 217 platforms