Anthropic published an updated privacy policy on June 8, 2026, with an effective date of July 8, 2026, introducing biometric identity verification, expanded third-party data sharing rules for agentic sessions, and new provisions covering research participation. The changes apply exclusively to consumer accounts - Claude Free, Pro, and Max plans - and mark the most substantive revision to the document since September 2025.

What changed and what did not

The previous version of the policy, which carried an effective date of January 12, 2026, remained in force until July 8. According to Anthropic's published change log, the January update had made two minor adjustments: adding a link to a new Consumer Health Data Privacy Policy and consolidating regional supplemental disclosures under a single section. The June 8 update is a different category of revision altogether.

Four areas were materially changed. The first covers multi-step tasks and connected apps. The second introduces verification data - the section at the center of most external commentary. The third adds a description of study participation data. The fourth expands the explanation of how Anthropic communicates with users and shares data with third parties.

According to Anthropic's summary of changes, the company "can increasingly carry out longer tasks and work with third-party apps and services" and has therefore "added details about the data involved when you use these features - what's shared with a third party when you connect a service, and what we receive in return when Claude completes tasks on your behalf."

Three commitments from the prior policy are carried forward without change, according to Anthropic: the company does not sell personal data, Claude remains ad-free, and users can control whether their conversations are used to improve Anthropic's AI models.

The biometric verification section in detail

The most technically significant addition is the Verification Data section in the updated policy document. According to Anthropic, "In certain circumstances, we may ask you to verify your age or identity." This provision is written in conditional language - "may ask" - but the data types specified are concrete.

The policy lists the following data categories depending on the verification method chosen: an image of a government-issued identity document and the information appearing on it, including the ID number and date of birth; a photo or video of the user's face; facial geometry templates, which the document flags "may be considered 'biometric data' in some jurisdictions"; and the outcome of the verification process, specifically whether age meets the applicable threshold.

Biometric data is a special category under the European Union's General Data Protection Regulation, carrying stricter legal requirements than ordinary personal data. Facial geometry templates are treated as biometric data under the Illinois Biometric Information Privacy Act in the United States and under a growing number of state-level frameworks. The policy does not specify which jurisdictions trigger biometric classification; it leaves that determination to the user or their legal counsel.

The verification mechanism is not operated by Anthropic itself. According to the policy, Anthropic works with service providers and business partners for purposes including "investigating and resolving safety and security issues, preventing and investigating fraud." Tech lawyer Tanya Chib, writing on LinkedIn, identified the identity verification provider as Persona Identities, a third-party KYC - Know Your Customer - firm. Her post, published shortly after the policy update circulated, noted that "most users will not notice that their biometric data is being processed by a company they have never heard of, under terms they have not separately reviewed."

Chib also flagged a GDPR concern: "The policy uses 'if you choose to do so', but it says nothing about what happens if you decline. Under GDPR, consent is only valid where refusal carries no detriment." The question of what users lose by declining verification is not addressed in the policy text.

The verification provision sits within the legal bases table in Section 10 of the policy. According to the document, for the purpose of preventing fraud, investigating abuse, and enforcing the Usage Policy, Anthropic may rely on consent "where you choose to verify your identity using biometric data," legitimate interests, legal obligation, and vital interests - the last of these being an EU GDPR basis typically reserved for life-or-death situations. The inclusion of vital interests alongside consent for a KYC procedure is unusual and is not explained in the accompanying change summary.

Agentic data flows: what the new text says

The expansion of the Inputs and Outputs section reflects the growing range of actions Claude can now take on behalf of users. According to the updated policy, interactions now include "agentic sessions (where Claude performs multi-step tasks or takes actions on your behalf), and connected services."

The prior policy described Inputs as covering chat, coding, and agentic sessions in general terms. The new text goes further. It states that "depending on the permissions you grant 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 your behalf." This is a formal acknowledgment that Claude's outputs are no longer confined to text displayed on a screen - they can modify external systems.

The third-party section of the policy was substantially rewritten to reflect this expanded footprint. According to Anthropic's updated policy, "Our Services allow you to connect to and interact with third-party applications, services, and content via the Services and through integrations you enable, such as Connectors, plugins, webhooks, and external APIs." Three specific scenarios are described. First, Claude may send Inputs, Outputs, and instructions to third-party services to perform actions on the user's behalf. Second, Claude may retrieve content and data from third-party services the user interacts with, and that data "becomes part of your Inputs." Third, some integrations may allow Claude to access third-party services "on an ongoing basis until you disable the feature or disconnect the integration in your settings."

The policy explicitly states that Anthropic "does not control, and is not responsible for, the data practices of Third-Party Services." Users are instructed to review each third party's privacy policy before enabling an integration. That obligation falls on the user, not on Anthropic.

The agentic scope of Claude has been building for some time. Anthropic launched its Integrations feature in May 2025, connecting Claude to remote MCP servers across web and desktop applications. In January 2026, the company released Cowork for macOS Max subscribers, giving Claude controlled file-system access alongside browser actions. By May 2026, the Claude Chrome Extension held 18.6% of measured agentic traffic, up from 17.3% the previous month. The updated privacy policy is, in part, a legal document catching up with product features already in users' hands.

Study participation data

A new category called Study Participation Data was added to the policy. According to Anthropic, "If you participate in Anthropic research studies, surveys, or interviews (for example our Anthropic Interviewer), we collect the responses you provide, associated account data, and related Technical Information. We may combine this data with other data from your account for aggregated analysis."

Study Participation Data is listed in the legal bases table under several processing purposes, including the provision of optional features, communications, security, and model training. It did not appear as a named data category in the January 2026 version of the policy.

Scope of application

The changes apply only to consumer accounts. According to Anthropic's notification to affected users, "These changes only affect consumer accounts (Claude Free, Pro, and Max plans). If you use Claude Team, Claude Enterprise, the Claude Platform, or other services under our Commercial Terms or other agreements, then these changes don't apply to you."

The definition of the Services was also narrowed slightly. The previous policy defined the Services as covering "our website and other places where Anthropic acts as a data controller" and explicitly included both consumer and commercial contexts. The updated version states the policy "does not apply to content that we process on behalf of customers of our business offerings, such as our Enterprise accounts."

Anthropic's European data controller is Anthropic Ireland, Limited, with a registered address at 6th Floor, South Bank House, Barrow Street, Dublin 4. Users outside the European Region deal with Anthropic PBC, registered at 548 Market St, PMB 90375, San Francisco, CA 94104. The data protection officer can be reached at dpo@anthropic.com.

Regional dimensions

The policy maintains supplemental disclosures for residents of Canada, Brazil, and the Republic of Korea. For Brazil, Anthropic relies on Standard Contractual Clauses approved by the Brazilian Data Protection Authority, the ANPD, for international data transfers. The Korean disclosure lists Anthropic Korea, Limited - represented by Patrick Azubike Ekeruo - with a registered address at 41F, 152 Teheran-ro, Gangnam-gu, Seoul, and the domestic data processing trustee as Bae, Kim and Lee LLC, reachable at +82-2-6252-2080.

The policy retains the provision that users in EU and UK member states can lodge complaints with their national supervisory authority. Australian users may contact the Office of the Australian Information Commissioner.

Data retention and deletion mechanics

The updated policy retains the 30-day back-end deletion window for individual conversations. According to the document, deleted conversations "will be removed immediately from your conversation history and automatically deleted from our back-end within 30 days." That period was unchanged from the prior version.

The policy added language on de-identification under a data minimization framing. According to Anthropic, "We apply aggregation and de-identification techniques where appropriate as part of our data minimization practices." Two specific scenarios are described. When users submit feedback, Anthropic dissociates Inputs and Outputs from user IDs before using them for model training. When the system flags content for potential policy violations, the content is dissociated from the user ID - though the policy notes that re-identification remains possible "to enforce our Terms of Service or Usage Policy with the responsible user if necessary."

The September 2025 policy update had introduced a five-year retention period for data used to train Claude, applying to chats and coding sessions where users had opted in. That retention period is not mentioned in the summary of changes published on June 8, suggesting it was not modified.

Context: Anthropic's data practices under scrutiny

The update arrives after a period of intense scrutiny of Anthropic's data practices across several dimensions. In February 2026, the Department of Justice argued in a federal filing that conversations with Claude do not carry attorney-client privilege, citing Anthropic's data collection practices and the possibility of government disclosure. In April 2026, researcher Alexander Hanff published findings that the Claude Desktop application had silently registered a Native Messaging bridge in seven Chromium-based browsers without user consent - a practice he categorized alongside Google Chrome's silent installation of Gemini Nano as a pattern of forced bundling across trust boundaries.

On the legal front, Anthropic reached a $1.5 billion settlement in September 2025 over the use of pirated books in model training, reported by PPC Land at the time as the largest publicly reported copyright recovery in history. Reddit filed a lawsuit against Anthropic on June 4, 2025, alleging unauthorized scraping of platform data to train Claude. The same crawling practices had been documented in detail when Anthropic updated its crawler documentation in February 2026.

For marketing professionals, the significance of the agentic data flow disclosures is direct. Claude is now integrated into a growing number of ad tech platforms. Meta opened its advertising infrastructure to Claude and other AI agents via MCP connectors in April 2026. DoubleVerify embedded MCP support allowing Claude to query media quality data. Lifesight launched an MCP connector in June 2026 giving Claude access to marketing measurement data for clients managing more than four billion dollars in combined spend. As these integrations deepen, the question of what data flows between Claude sessions and third-party ad infrastructure becomes commercially relevant - not just for consumer privacy, but for the handling of advertiser data entering Claude-powered workflows.

The policy explicitly notes that users are "responsible for ensuring you have the necessary rights, permissions, or authority" when providing Inputs - including when uploading files, granting Claude access to third-party services, or instructing Claude to retrieve or act on information.

Timeline

  • May 1, 2025 - Anthropic launched the Integrations feature, connecting Claude to remote MCP servers and expanding its access to third-party tools and data
  • August 26, 2025 - Anthropic launched Claude for Chrome as a research preview with 1,000 Max plan users, extending Claude's ability to take real-world actions inside a browser
  • September 5, 2025 - Anthropic agreed to a $1.5 billion settlement in a copyright lawsuit over pirated books used in model training
  • September 28, 2025 - Anthropic updated its privacy policy to introduce a five-year retention period for model training data and to add Technical Information to the legal bases table
  • January 12, 2026 - Anthropic updated the privacy policy to add a link to the Consumer Health Data Privacy Policy and consolidate regional supplemental disclosures; Anthropic also released Cowork for macOS Max subscribers
  • February 6, 2026 - The Department of Justice filed a motion arguing Claude conversations do not carry attorney-client privilege, citing Anthropic's data collection and disclosure practices
  • February 25, 2026 - Anthropic updated documentation for its three web crawlers - ClaudeBot, Claude-User, and Claude-SearchBot - specifying the consequences of blocking each
  • April 18, 2026 - Researcher Alexander Hanff published findings that Claude Desktop had silently registered a Native Messaging bridge in seven Chromium-based browsers without user consent
  • April 29, 2026 - Meta launched open beta of Meta Ads AI Connectors, giving Claude and ChatGPT access to advertiser accounts via MCP
  • June 4, 2025 - Reddit filed a lawsuit against Anthropic alleging unauthorized scraping of platform data to train Claude
  • June 8, 2026 - Anthropic published the updated privacy policy with an effective date of July 8, 2026, introducing biometric verification data, agentic data flow disclosures, and study participation provisions
  • July 8, 2026 - The updated Anthropic privacy policy becomes effective for all consumer accounts (Claude Free, Pro, and Max)

Summary

Who - Anthropic PBC, the AI safety and research company behind the Claude AI assistant, published the update. It affects all consumer accounts: Claude Free, Pro, and Max plan subscribers worldwide. Commercial accounts - Claude Team, Claude Enterprise, and users of the Claude Platform under commercial agreements - are not covered by these changes.

What - Anthropic updated its privacy policy to add four new areas: biometric verification data collection through a third-party KYC provider; expanded disclosure of data flows in agentic and multi-step task sessions; a new study participation data category; and additional information about how Anthropic communicates with users and shares data with third parties. The core commitments - no data sales, no advertising in Claude, and user control over model training consent - were retained unchanged.

When - The updated policy was published on June 8, 2026. It becomes effective on July 8, 2026. The previous version, effective January 12, 2026, remains in force until that date.

Where - The policy governs use of Claude.ai, the Anthropic website, and other Anthropic consumer products globally. The European data controller is Anthropic Ireland, Limited, based in Dublin. The US data controller is Anthropic PBC, based in San Francisco. Regional supplemental disclosures apply to users in Canada, Brazil, and the Republic of Korea.

Why - According to Anthropic, the update reflects changes in what Claude can do: longer multi-step tasks, connections to external apps and services, and a broader set of interactions that generate data flows to and from third parties. The verification section addresses age and identity checks introduced as a safety measure. The expansion of data flow disclosures aligns the legal document with product features - including agentic sessions, Connectors, plugins, webhooks, and external APIs - that were already available to users before the policy was updated.