Anthropic AI Footnotes
What Anthropic's AI features do with your data, as Anthropic publishes it. Checked October 31, 2026. Anthropic in the Directory.
Agent access
Claude API at https://api.anthropic.com for programmatic access to Claude models and Claude Managed Agents, authenticated with an API key or a Workload Identity Federation token. source · checked October 31, 2026
Claude for Enterprise
Claude Enterprise plan
| On by default? | Not published as of October 31, 2026 |
|---|---|
| How an admin turns it off | Owners can stop members from sending thumbs up/down feedback, the main path by which chats can enter training, with the Rate chats setting in Organization settings > Data and Privacy. source · checked October 31, 2026 |
| Trains on customer data? | No training on inputs or outputs by default. Chats can be used for training if a user reports feedback or bugs (for example, thumbs up/down) or the customer otherwise allows it. source · checked October 31, 2026 |
| Retention | Retained indefinitely unless an Owner sets a custom retention period (minimum 30 days) in Organization settings > Data and Privacy. Deleted chats are removed from back-end storage within 30 days; chats flagged for Usage Policy violations are kept up to 2 years. source, source 2, source 3, source 4 · checked October 31, 2026 |
| Where data is processed | Data is stored in the US. On usage-based Enterprise plans, an admin can turn on US-only inference in Organization settings > Data and privacy, billed at 1.1x API rates for Opus 4.6, Sonnet 4.6 and later models. source, source 2 · checked October 31, 2026 |
| Model providers | Not published as of October 31, 2026 |
| Price | $20 per seat per month billed annually, 20-seat minimum, plus usage billed at API rates. source · checked October 31, 2026 |
Claude Team
Claude Team plan
| On by default? | Not published as of October 31, 2026 |
|---|---|
| How an admin turns it off | Not published as of October 31, 2026 |
| Trains on customer data? | No training on inputs or outputs by default; feedback submissions can be used for training, and Owners can turn feedback off. source · checked October 31, 2026 |
| Retention | Chats are kept in the product for history. A deleted chat leaves chat history immediately and is deleted from back-end storage within 30 days. Custom retention controls are an Enterprise feature, not Team. source · checked October 31, 2026 |
| Where data is processed | Data is stored in the US. Traffic may be routed to select countries in the US, Europe, Asia and Australia by default. source · checked October 31, 2026 |
| Model providers | Not published as of October 31, 2026 |
| Price | Per seat. Standard seat: $20 per seat per month billed annually, $25 billed monthly. Premium seat: $100 billed annually, $125 billed monthly. For teams of 2 to 150. source · checked October 31, 2026 |
Claude Code
Claude Team and Enterprise plans
| On by default? | Not published as of October 31, 2026 |
|---|---|
| How an admin turns it off | Access follows the member's seat. On older Enterprise plans, Owners assign seats that include Claude Code in Organization settings > Organization. source · checked October 31, 2026 |
| Trains on customer data? | Anthropic does not train generative models on code or prompts sent to Claude Code under commercial terms (Team, Enterprise, API), unless the customer opts in, for example through the Development Partner Program. source · checked October 31, 2026 |
| Retention | Commercial standard: 30 days. Zero data retention is available to qualified Claude for Enterprise accounts, enabled per organization by Anthropic; it is not in the standard Enterprise plan. Local transcripts are kept on the machine for 30 days by default. source · checked October 31, 2026 |
| Where data is processed | US-only inference, where turned on for a usage-based Enterprise plan, covers Claude Code. source · checked October 31, 2026 |
| Model providers | Not published as of October 31, 2026 |
| Price | Included with every Team seat and with the Enterprise seat on new and self-serve Enterprise plans. source · checked October 31, 2026 |
Connectors (remote MCP)
Claude Team and Enterprise plans
| On by default? | Off until an Owner or Primary Owner enables a connector for the organization; each member then authenticates individually. source · checked October 31, 2026 |
|---|---|
| How an admin turns it off | Organization settings > Connectors. On Team, only Owners and Primary Owners add custom connectors; on Enterprise, also custom roles with Manage access to Libraries. source · checked October 31, 2026 |
| Trains on customer data? | Not published as of October 31, 2026 |
| Retention | Not published as of October 31, 2026 |
| Where data is processed | Claude reaches remote MCP servers from Anthropic's cloud infrastructure. Connected services process data on their own infrastructure, which may be outside the US; the US-only inference setting does not change that. source, source 2 · checked October 31, 2026 |
| Model providers | Not published as of October 31, 2026 |
| Price | Included in Team and Enterprise plans, with admin controls for remote and local connectors. source · checked October 31, 2026 |
Claude API
Claude Platform (Claude API)
| On by default? | Not published as of October 31, 2026 |
|---|---|
| How an admin turns it off | Not published as of October 31, 2026 |
| Trains on customer data? | No training on API inputs or outputs by default. source · checked October 31, 2026 |
| Retention | Inputs and outputs are deleted from the backend within 30 days, except for features with customer-controlled storage (such as the Files API), zero data retention agreements, Usage Policy enforcement or legal requirements. Zero data retention is available on request through sales, per organization. source · checked October 31, 2026 |
| Where data is processed | inference_geo set to "us" or "global" per request or as a workspace default; US-only inference costs 1.1x on Claude 4.6 and later models. Workspace geo (storage) is currently US only. source · checked October 31, 2026 |
| Model providers | Not published as of October 31, 2026 |
| Price | Consumption, per million tokens by model. Claude Opus 5.5 is billed at $4 per million input tokens and $20 per million output tokens. source · checked October 31, 2026 |
Connections to outside MCP servers
Claude connects to outside remote MCP servers as custom connectors on Team and Enterprise. An Owner or Primary Owner adds them in Organization settings > Connectors (on Enterprise, also custom roles with Manage access to Libraries), and each member then authenticates. Owners can restrict each connector's actions org-wide. No stage label is published for connectors; installing connectors on mobile is in beta. source, source 2, source 3, source 4 · checked October 31, 2026
Compliance
- HIPAA: BAA for HIPAA-ready services: the first-party API, Claude in Microsoft Foundry and Enterprise plans. Team has no HIPAA-ready offering. On Enterprise, coverage starts only after the Primary Owner turns on HIPAA compliance and accepts the BAA. Covered: chat, projects and artifacts in chat, file creation and code execution (no network access), voice, web search and research. Not covered: skills and data sent to third parties through connectors, Enterprise Search or Claude in Chrome. Excluded: Claude Console, Cowork in the cloud and beta features such as Claude in Office, Claude Design, Slides and Docs. Claude Code in the terminal and the Code tab is covered once the HIPAA configuration is applied. On the API, eligible organizations sign the BAA and enable HIPAA readiness in the Claude Console; not all API features are covered. source, source 2, source 3, source 4, source 5 · checked October 31, 2026
- FedRAMP: FedRAMP High for Claude for Government, and for Claude on Amazon Bedrock in AWS GovCloud and on Google Vertex AI in Assured Workloads as partner hosting environments. The Trust Center lists FedRAMP High as N/A for the Claude API, Claude Enterprise and Claude Team. source, source 2 · checked October 31, 2026
EU AI Act
Anthropic states it has signed the EU AI Act's Article 50(2) Code of Practice on Transparency of AI-Generated Content. On July 21, 2025, it stated it intends to sign the General-Purpose AI Code of Practice. The European Commission lists Anthropic among the signatories of the General-Purpose AI Code of Practice (as of October 31, 2026). source, source 2, source 3 · checked October 31, 2026
Not published by Anthropic as of October 31, 2026
- Data storage regions other than the US for Claude Team and Enterprise
- A stage label (beta or generally available) for custom remote MCP connectors in Team and Enterprise
Changes
No changes recorded since the first check on October 31, 2026.
How this page is made, and how to send a correction: method.