Anthropic · Anthropic Privacy Policy (Superseded Capture) · View original document ↗

Model Training Opt-Out with Safety-Review Carve-Out

Medium severity Medium confidence Explicitdocumentlanguage Common · 222 of 352 platforms
Get alerted the next time Anthropic changes these terms. Get same-day alerts →
Share 𝕏 Share in Share 🔒 PDF
Recent governance activity Anthropic recorded 3 documented changes in the last 30 days.
Get same-day alerts →
Monitor governance changes for Anthropic Monitor emails you the same day this changes. The archive stays free.
Get same-day alerts →

Get the weekly 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. No account.

Document Record

What it is

The agreement permits use of user Inputs and Outputs for AI model training by default, with an opt-out available in account settings, but specifies two conditions under which training use continues regardless of opt-out status: when conversations are flagged for safety review, and when users have explicitly submitted content as feedback.

This analysis describes what Anthropic's agreement states, permits, or reserves. It does not constitute a legal determination about enforceability. Regulatory applicability and practical outcomes may vary by jurisdiction, enforcement context, and individual circumstances. Read our methodology

ConductAtlas Analysis

Why it matters (compliance & governance perspective)

This provision establishes a conditional opt-out structure in which the training data opt-out right is subject to two categorical exceptions that may encompass a meaningful subset of user conversations. Compliance teams should evaluate whether this carve-out structure satisfies the opt-out right requirements under CCPA and the purpose limitation and legal basis requirements under GDPR and UK GDPR.

Interpretive note: The scope of what qualifies as safety-flagged content is not defined in the policy, creating ambiguity about how broadly the carve-out may apply in practice.

Clause Stability Stable

0
Changes
4
Months Monitored
May 10, 2026
First Seen
Jul 9, 2026
Last Seen
This clause type exists across 949 other provisions on other platforms.

Consumer impact (what this means for users)

Under this clause, users can limit use of their conversations for model training through account settings, but the agreement retains authorization to use safety-flagged or explicitly reported conversations for training purposes regardless of that setting. The specific scope of what qualifies as safety-flagged content is not defined with precision in the policy.

What you can do

⚠️ These actions may provide transparency or partial mitigation but may not fully address the underlying issue. Effectiveness varies by jurisdiction and individual circumstances.
  • Opt Out of Arbitration
    Log in to your Anthropic account, navigate to Privacy Settings, and disable the option allowing your Inputs and Outputs to be used for model training.

How other platforms handle this

Microsoft Medium

Microsoft also emphasizes the importance of validating AI models responsibly to enhance fairness and alignment with reality.

Segment Medium

training AI/ML models with performance metrics to optimize network reliability; providing dedicated customer support; and, refining our Service suite through usage insights.

Tinder Medium

This is still Your Content, and you are responsible for it and its accuracy, as well as your use of it on our Services and any and all decisions made, actions taken, and failures to take action based on Your Content.

See all platforms with this clause type →

Monitoring

Anthropic has changed this document before.

Receive same-day alerts, structured change summaries, and monitoring for up to 25 platforms.

Get Monitor Or create a free account →
▸ View Original Clause Language DOCUMENT RECORD
"
We may use your Inputs and Outputs to train and improve Anthropic AI models, unless you opt out through your account settings. Even if you opt-out, we will use Inputs and Outputs for model improvement when: (i) your conversations are flagged for safety review to improve our ability to detect harmful content, enforce our policies, or advance AI safety research, or (ii) you've explicitly reported the materials to us (for example via our feedback mechanisms).

Excerpt from Anthropic's Privacy Policy (Superseded Capture)

ConductAtlas Analysis

Institutional analysis (regulatory & governance intelligence)

REGULATORY LANDSCAPE: This provision implicates CCPA's opt-out of sale and sharing rights, GDPR and UK GDPR purpose limitation and legal basis requirements (Articles 5 and 6), and the FTC Act's prohibition on unfair or deceptive practices. The carve-out for safety-flagged content invokes a legitimate interest basis that EU and UK data protection authorities may scrutinize for proportionality. The FTC is the primary US enforcement authority. GOVERNANCE EXPOSURE: High. The safety-review carve-out creates potential tension with CCPA opt-out rights if safety flagging is applied broadly, and with GDPR purpose limitation if training use of flagged data is not clearly disclosed as a compatible purpose. The scope of what constitutes safety-flagged content is not defined in the policy, creating interpretive uncertainty about how broadly this exception may apply. JURISDICTION FLAGS: California (CCPA opt-out of sharing), EU and UK (GDPR purpose limitation and legitimate interest balancing), Brazil (LGPD consent and legitimate interest grounds). The policy's separate legal bases table lists Scientific Research and Legitimate Interests for model training, which may face challenge in jurisdictions requiring explicit consent for AI training. CONTRACT AND VENDOR IMPLICATIONS: Enterprise and B2B customers using separate agreements are excluded from this policy, but organizations whose employees use consumer accounts should assess whether training data flows from consumer accounts affect confidentiality or data protection obligations. Procurement teams integrating Claude via API under consumer terms should verify which agreement governs their data. COMPLIANCE CONSIDERATIONS: Legal teams should assess whether the safety-review carve-out is sufficiently specific to satisfy transparency requirements under GDPR and CCPA. A consent mechanism audit should evaluate whether the opt-out is clearly presented at or before first use. Data mapping should document the distinction between opted-out user data used under the carve-out versus standard training data flows.

Full institutional analysis

Regulatory citations, enforcement risk, and due diligence action items.

Get same-day alerts when this changes → Get Analyst

Monitor: same-day alerts on the platforms you choose. Analyst: full institutional analysis.

Applicable agencies

  • FTC
    The FTC has jurisdiction over consumer data practices including opt-out mechanisms and representations about data use under the FTC Act.
    File a complaint →

Applicable regulations

EU AI Act
European Union
California AB 2013 AI Training Data Transparency
US-CA
Colorado AI Act
US-CO
EU AI Act - High Risk Provisions
EU
GDPR
European Union
Texas AI Act
Texas, USA
Trump Executive Order on AI Policy Framework
US
UK GDPR
United Kingdom

Provision details

Document information
Document
Anthropic Privacy Policy (Superseded Capture)
Entity
Anthropic
Document last updated
May 5, 2026
Tracking information
First tracked
July 9, 2026
Last verified
July 9, 2026
Record ID
CA-P-008335
Document ID
CA-D-00012
Evidence Provenance
Source URL
Wayback Machine
Content hash (SHA-256)
e91b78d120f18b8a635385fb036a9ad6b0135fe530a2e4aadcc4d575da32fca0
Analysis generated
July 9, 2026 17:12 UTC
Methodology
Evidence
✓ Snapshot stored   ✓ Hash verified
Citation Record
Entity: Anthropic
Document: Anthropic Privacy Policy (Superseded Capture)
Record ID: CA-P-008335
Captured: 2026-07-09 17:12:50 UTC
SHA-256: e91b78d120f18b8a…
URL: https://conductatlas.com/platform/anthropic/anthropic-privacy-policy-superseded-capture/provision/CA-P-008335/model-training-opt-out-with-safety-review-carve-out/
Accessed: July 23, 2026
Permanent archival reference. Stable identifier suitable for legal filings, compliance documentation, and research citation.
Classification
Severity
Medium
Categories

Other risks in this policy

Related Analysis

Compliance Governance Intelligence

Need to monitor specific governance provisions?

Compliance includes provision-level monitoring, governance timelines, regulatory mapping, and audit-ready analysis.

Arbitration clauses AI governance Data rights Indemnification Retention policies
Get Compliance

Or start with Monitor →

Built from archived source documents, structured governance mappings, and historical version tracking.

Frequently Asked Questions

What does Anthropic's Model Training Opt-Out with Safety-Review Carve-Out clause do?

This provision establishes a conditional opt-out structure in which the training data opt-out right is subject to two categorical exceptions that may encompass a meaningful subset of user conversations. Compliance teams should evaluate whether this carve-out structure satisfies the opt-out right requirements under CCPA and the purpose limitation and legal basis requirements under GDPR and UK GDPR.

How does this clause affect you?

Under this clause, users can limit use of their conversations for model training through account settings, but the agreement retains authorization to use safety-flagged or explicitly reported conversations for training purposes regardless of that setting. The specific scope of what qualifies as safety-flagged content is not defined with precision in the policy.

How many platforms have this type of clause?

ConductAtlas has identified this type of provision across 222 platforms. See the full comparison.

Is ConductAtlas affiliated with Anthropic?

No. ConductAtlas is an independent monitoring service. We are not affiliated with, endorsed by, or sponsored by Anthropic.