Anthropic may train on customer data in Development Partner Mode.
Anthropic’s Service Specific Terms say that if a customer enables Development Partner Mode, Anthropic may use the data the customer submits, including to train models.
In September 2026, Anthropic added a Republic of Korea addendum to its Privacy Policy and published transparency disclosures for two new model variants.
We monitor 37 Anthropic documents, including its Commercial and Consumer Terms of Service, Privacy Policy, Data Processing Addendum and Usage Policy.
What Anthropic’s documents say about client data.
Quoted from the current version of each document. The right-hand column shows how many of the 75 vendors we monitor have a matching clause.
Using customer data to train AI
Addressed“If Customer enables Development Partner Mode, Anthropic may use the data that Customer submits to the Services (e.g., Customer Content) in connection with Anthropic's products and services, including to train models.”
Sharing data with third parties
Addressed“We may comply with governmental, court, and law enforcement requests or requirements relating to provision or use of the Services, or to information provided to or collected under our Terms.”
What changed, and when.
Notable changes since monitoring began in February 2026, newest first. Each is written from our automated comparison of the before and after versions, then reviewed by a person.
Added transparency disclosures for Claude Sonnet 5.5
Lists the model’s release and modalities, and says its training mix may include user data from feedback or bug reports.
Added transparency disclosures for Claude Fable 5.1 and Mythos 5.1
Covers how each variant is deployed and classified for risk, and the mix of data each was trained on.
Added a Republic of Korea addendum to the Privacy Policy
Applies where Anthropic is subject to Korea’s PIPA. Lists data collected for identity verification and fraud monitoring, kept until account deletion.
Added terms for trusted access and verification programs
Customers can apply for programs that unlock certain models or features. Any program terms must be accepted before access.
Added deletion and student data terms to its US K-12 DPA
Anthropic does not collect student data directly from students, and will delete customer or student data on request within a reasonable time.
+ 171 other updates detected, such as formatting, page layout and wording changes.
- Oct 8, 2026: Privacy Policy
- Oct 6, 2026: Usage Policy
- Oct 6, 2026: Consumer Health Data Privacy Policy
- Oct 6, 2026: Terms of Service (US K-12)
- Oct 5, 2026: Terms of Service: Commercial
- Oct 4, 2026: Public Data Crawl Notice
- Oct 3, 2026: Supplemental Credits Terms
- Oct 2, 2026: Privacy Policy
37 Anthropic documents.
Sub-documents are other Anthropic documents that its policies link to, such as addenda and product terms. We list the ones identified so far; there may be others.
Anthropic terms: common questions.
Does Anthropic train models on business customer data?
Anthropic’s Service Specific Terms say that if a customer enables Development Partner Mode, Anthropic may use the data the customer submits, including to train models.
When did Anthropic last change its privacy policy?
On September 12, 2026, Anthropic added a Republic of Korea addendum to its Privacy Policy, covering users where Korea’s PIPA applies.
Which Anthropic documents does Legal Sentinel monitor?
37 public documents, including the Commercial and Consumer Terms of Service, Privacy Policy, Data Processing Addendum, Usage Policy and Transparency Policy.