Covered Models (original) (raw)

Anthropic may designate certain models as “Covered Models” when they cross capability thresholds that warrant additional safeguards or other treatment. This page lists the models currently designated as Covered Models and describes the data handling, privacy, and access policies that apply to them.

What is a Covered Model?

A Covered Model is a Claude model whose capabilities—for example in software engineering, agentic workflows, scientific reasoning, or cybersecurity — represent a substantial step up from prior generations and create elevated risk if misused. Because some forms of misuse only become detectable across many requests, these models require safeguards that operate over a retained window of usage data rather than on a single request at a time.

Anthropic may designate certain models as "Covered Models" when their capabilities warrant additional safeguards or other treatment. When a model is designated as a Covered Model, it will be added to the list below and the policies on this page take effect for that model on every surface where it is offered.

Current Covered Models

We will update this list as new models are designated or as existing designations change.

Policies that apply to Covered Models

The following policies apply to every Covered Model listed above, on every platform where it is available (Claude apps, Claude Platform, Amazon Bedrock, Google Cloud Vertex AI, and Microsoft Foundry).

Data retention

Use and review of retained data

Security and privacy controls

Availability and access

Business Associate Agreements (BAA) for Commercial Customers HIPAA-ready Enterprise plans Public Sector FAQs Data retention practices for Mythos-class models Covered Models under a Business Associate Agreement (BAA)