Data retention - verified
Does Anthropic store your API prompts?
Default retention: within 30 days, or not retained by default. Trains on your API data: No. Zero data retention: by approval (by arrangement with Anthropic sales, per organization). Regional processing: US. Anthropic's own pages disagree on default retention; both versions are quoted below.
Scope: Claude API (api.anthropic.com), Claude Platform on AWS, Claude in Microsoft Foundry where Anthropic is processor. Not Claude.ai consumer plans. On Bedrock/Google Cloud the cloud provider is processor.
Read from 4 official sites. Compare every AI API's data retention
- Trains on data
- No
- ZDR
- By approval
- Regions
- US
- Verified
- 2026-10-18
The docs disagree
Anthropic's own pages say different things here. Both are quoted; neither is picked. Ask Anthropic which applies to your account.
The docs disagree
Default retention
Anthropic's privacy center says API inputs and outputs are deleted within 30 days of receipt or generation. Its API data retention doc says conversation content is not retained by default, except for Covered Models, which require 30-day retention. Both pages are Anthropic's own; we show both and do not pick one.
“For Anthropic API users, we automatically delete inputs and outputs on our backend within 30 days of receipt or generation, except:”
“Only what is technically necessary for the feature to work is retained. Conversation content (your prompts and Claude's outputs) is not retained by default; the exception is Covered Models, which require 30-day retention.”
How long are prompts kept by default?
- Default retention
Within 30 days, or not retained by default
Privacy Center: API inputs and outputs automatically deleted within 30 days of receipt or generation, with exceptions (Files API, ZDR agreements, Usage Policy enforcement, law). Flagged content up to 2 years. Covered Models (Fable/Mythos class) require 30-day retention even for ZDR orgs.
“For Anthropic API users, we automatically delete inputs and outputs on our backend within 30 days of receipt or generation, except:”
privacy.claude.com “We retain inputs and outputs for up to 2 years and trust and safety classification scores for up to 7 years if your chat is flagged by our automated trust and safety systems as violating our Usage Policy.”
privacy.claude.com “Only what is technically necessary for the feature to work is retained. Conversation content (your prompts and Claude's outputs) is not retained by default; the exception is Covered Models, which require 30-day retention.”
platform.claude.com - Duration
within 30 days
- Purpose
Not stated as a single purpose; exceptions for Usage Policy enforcement and law
Does it train on your API data?
No
Commercial Terms bar training on customer content.
“Anthropic may not train models on Customer Content from Services.”
“Retained data is never used for model training without your express permission.”
Zero data retention (ZDR)
- Availability
By approval: the provider must approve your request
By arrangement with Anthropic sales, per organization
- Who can get it
By arrangement via Anthropic sales; enabled per organization (each new org separately). Covers Messages and Token Counting APIs for eligible features, Claude Code with Commercial-org API keys, Claude Platform on AWS on request.
- Not covered
Console/playground, Managed Agents, Batch (29 days), Files API, Code execution and programmatic tool calling (up to 30 days), Agent skills, MCP connector, MCP tunnels, Claude for Excel, consumer plans, Teams/Enterprise UIs (separate Claude Code Enterprise ZDR), CORS not supported, Covered Models (Fable 5/5.1, Mythos 5/5.1) unless expressly authorized; flagged content and legal holds.
- From the docs
“Under a ZDR arrangement, Anthropic does not store customer prompts or responses at rest after the API response is returned. To request ZDR for your organization, contact the Anthropic sales team. ZDR is enabled per organization; each new organization requires ZDR to be enabled separately by your account team, and enablement does not automatically extend to other organizations under the same account.”
platform.claude.com “Claude Fable 5.1, Claude Mythos 5.1, Claude Fable 5, and Claude Mythos 5: These models require 30-day data retention and are not available under ZDR unless expressly authorized by Anthropic.”
platform.claude.com “Even with ZDR or HIPAA arrangements in place, Anthropic may retain data where required by law or where it has been flagged by Anthropic's automated trust and safety systems. As a result, if a chat or session is flagged, Anthropic may retain inputs and outputs for up to 2 years.”
platform.claude.com
Features kept longer
Stateful features hold your content until a job finishes, a window expires or you delete it, often regardless of the default above.
| Feature | Retention | The docs say |
|---|---|---|
| Batch processing (/v1/messages/batches) | up to 29 days after batch creation | “Batch processing stores request and response data for up to 29 days after batch creation.”platform.claude.com |
| Files API (/v1/files) | Until deleted or expires_at (optional 1 hour to 90 days); metadata visible up to 30 days after expiry; content may be retained a limited period for safety review | “Files persist until you delete them with the DELETE /v1/files/{file_id} endpoint or they reach their expires_at”platform.claude.com |
| Files API expiration window | expires_in_seconds 3,600 to 7,776,000 (1 hour to 90 days) | “The value is an integer number of seconds between 3,600 (1 hour) and 7,776,000 (90 days).”platform.claude.com |
| Code execution and programmatic tool calling | Container data up to 30 days | “Container data retained up to 30 days. See Code execution.”platform.claude.com |
| Structured outputs / strict tool use | JSON schema cached up to 24 hours since last use (prompts/outputs not stored) | “Only the JSON schema is cached, for up to 24 hours since last use.”platform.claude.com |
| Prompt caching | In memory for cache TTL, then promptly deleted | “KV cache representations and cryptographic hashes are held in memory for the cache TTL and promptly deleted after expiry.”platform.claude.com |
| Claude Managed Agents sessions | Until you delete them | “Sessions are stateful resources; transcripts persist until you delete them.”platform.claude.com |
| Agent skills, MCP connector | Per standard policy | “Skill data retained per standard policy.”platform.claude.com |
| Covered Models (Fable 5/5.1, Mythos 5/5.1) | 30 days, on every platform where offered | “Prompts submitted to, and outputs generated by, covered models are retained for 30 days to support our safety work, on every platform where these models are offered.”privacy.claude.com |
| Feedback (thumbs up/down, bug reports) | 5 years | “Where you have provided feedback to us (e.g. by submitting feedback through our thumbs up/down button or sent bug reports), we retain data associated with that submission for 5 years.”privacy.claude.com |
Where is your data stored and processed?
- Regional processing
US
US-only inference with inference_geo, Claude 4.6 and later
- Stated location
Global inference by default
- Regions
- us (inference geo and workspace geo)
- global (inference geo, default)
- How to choose
inference_geo request parameter ("us" or "global") on POST /v1/messages, or workspace default_inference_geo / allowed_inference_geos (Console or Admin API data_residency field). Workspace geo chosen at workspace creation, cannot be changed; only "us" available. Supported on Claude 4.6 and later models; older models return 400. US-only inference priced at 1.1x.
- What it covers
Two independent settings: inference geo controls where model inference runs; workspace geo controls storage at rest and endpoint processing (image transcoding, code execution).
- From the docs
“Inference geo: Controls where model inference runs, on a per-request basis. Set through the inference_geo API parameter or as a workspace default.”
platform.claude.com “Workspace geo: Controls where data is stored at rest and where endpoint processing (such as image transcoding and code execution) happens. Configured at the workspace level in the Claude Console.”
platform.claude.com “Workspace geo is set when you create a workspace and can't be changed afterward. Currently, "us" is the only available workspace geo.”
platform.claude.com “Inference geo: Only "us" and "global" are available.”
platform.claude.com “Claude 4.6 and later models: US-only inference (inference_geo: "us") is priced at 1.1x the standard rate across all token pricing categories (input tokens, output tokens, cache writes, and cache reads).”
platform.claude.com
Regional endpoints
Regional hostnames: Not stated
No regional hostnames documented; residency is selected with the inference_geo parameter on api.anthropic.com. inference_geo is not available on Bedrock, Google Cloud, Foundry, or the OpenAI SDK compatibility endpoint.
Certifications
- SOC 2 Type 2
- ISO 27001
- ISO 42001
- CSA STAR
- HIPAA (BAA; HIPAA-ready API self-serve in Console)
- NIST 800-171
“Eligible organizations can review and execute the BAA and enable HIPAA readiness directly from the Claude Console.”
Research notes
- HIPAA readiness is an alternative to ZDR for PHI; it is not available on Claude Platform on AWS or Microsoft Foundry.
- Covered Models policy effective June 9, 2026; ZDR orgs can enable 30-day retention per workspace to use them.
- Legacy org-level US-only opt-out was migrated to allowed_inference_geos: ["us"], default_inference_geo: "us".
- Trust center certification row was read from a rendered table at https://trust.anthropic.com/ with check marks under SOC 2 Type 2, ISO 27001, ISO 42001, CSA Star, HIPAA and NIST 800-171, and N/A under FedRAMP High, DoD IL4 and DoD IL5.
FAQ
Does Anthropic store your API prompts?
Within 30 days, or not retained by default. Privacy Center: API inputs and outputs automatically deleted within 30 days of receipt or generation, with exceptions (Files API, ZDR agreements, Usage Policy enforcement, law). Flagged content up to 2 years. Covered Models (Fable/Mythos class) require 30-day retention even for ZDR orgs. The provider's own pages disagree. Privacy center: "For Anthropic API users, we automatically delete inputs and outputs on our backend within 30 days of receipt or generation, except:" API and data retention docs: "Only what is technically necessary for the feature to work is retained. Conversation content (your prompts and Claude's outputs) is not retained by default; the exception is Covered Models, which require 30-day retention."
Does Anthropic train on your API data?
No. The provider says: "Anthropic may not train models on Customer Content from Services."
Does Anthropic offer zero data retention?
By approval (by arrangement with Anthropic sales, per organization). By arrangement via Anthropic sales; enabled per organization (each new org separately). Covers Messages and Token Counting APIs for eligible features, Claude Code with Commercial-org API keys, Claude Platform on AWS on request.
Where does Anthropic process your data?
Regional processing options: US (US-only inference with inference_geo, Claude 4.6 and later). Stated processing location: Global inference by default.
Sources
Last verified from Anthropic's own pages below. Quotes are verbatim.
- https://platform.claude.com/docs/en/manage-claude/api-and-data-retention
- https://platform.claude.com/docs/en/manage-claude/data-residency
- https://privacy.claude.com/en/articles/7996866-how-long-do-you-store-my-organization-s-data
- https://privacy.claude.com/en/articles/15425996-data-retention-practices-for-covered-models
- https://www.anthropic.com/legal/commercial-terms
- https://platform.claude.com/docs/en/build-with-claude/files
- https://platform.claude.com/docs/en/build-with-claude/batch-processing
- https://trust.anthropic.com/
This page summarizes what each provider's own pages say. It is not legal advice; your contract, data processing addendum and the provider's current terms govern what happens to your data.
Newsletter
Know when a provider changes its data terms
One email per week. New model releases, pricing and policy changes, and what they mean for what you build.