Claude Zero Data Retention (ZDR): Enterprise API Guide
What is Claude Zero Data Retention (ZDR)?
Claude's Zero Data Retention (ZDR) means Anthropic does not store your API inputs or outputs after the response is returned. ZDR is available only to eligible API enterprise customers by separate agreement — not through claude.ai or standard plans.
- ZDR eliminates post-response storage of inputs and outputs — exceptions apply for safety classifiers and flagged violations.
- ZDR requires a separate agreement with Anthropic and is not enabled by default even on Enterprise plans.
- Claude.ai Free, Pro, and Team plans are not eligible for ZDR — API enterprise deployment only.
- ZDR does not replace the need for a DPA or SCCs — both remain required under GDPR.
Claude’s Zero Data Retention (ZDR) means Anthropic does not store your API inputs or outputs after the response is returned. ZDR is available to eligible API enterprise customers through a separate agreement with Anthropic — it is not a default setting, not available on claude.ai consumer or Team plans, and not automatically included in a Claude Enterprise purchase. For enterprise teams in Germany that need to meet GDPR data minimisation requirements, ZDR is one of the key levers in the Claude compliance setup — alongside the DPA, SCCs, and optionally EU data residency.
This article explains what ZDR covers, who is eligible, what it does not cover, and how it fits into a broader GDPR compliance framework for companies deploying Claude via the API.
What Is Claude Zero Data Retention (ZDR)?
Zero Data Retention (ZDR) is a contractual commitment from Anthropic that API inputs and outputs are not stored after the API response is returned to the caller. Under standard API terms without ZDR, Anthropic retains inputs and outputs for a period defined in the DPA and privacy policy — primarily for trust-and-safety review and product improvement. Under ZDR, that retention is eliminated for covered content.
The practical effect for enterprise deployments: conversations, documents, or prompts submitted to the Claude API are processed to generate a response and then discarded — they are not retained in Anthropic’s systems beyond the immediate inference cycle.
What ZDR is not:
- It is not an EU data residency commitment — processing still occurs in Anthropic’s infrastructure, primarily in the United States, unless you deploy via AWS Bedrock or Google Vertex AI EU regions
- It is not a replacement for the DPA or SCCs — those govern the legal framework for personal data processing, regardless of retention duration
- It is not a guarantee that no data is transmitted — inputs are still sent to and processed by Anthropic’s systems; ZDR only governs post-response storage
For a complete picture of Claude’s data privacy setup for enterprise customers, see our Claude Enterprise GDPR compliance guide.
Who Is Eligible for Claude ZDR?
ZDR is available to eligible API enterprise customers who meet Anthropic’s requirements and have received explicit approval. The eligibility framework is:
| Customer type | ZDR available? |
|---|---|
| claude.ai Free plan | No |
| claude.ai Pro plan | No |
| Claude.ai Team plan | No |
| Claude Enterprise (claude.ai) | Not by default — requires separate ZDR agreement |
| API enterprise customers (commercial agreement) | Yes, with Anthropic approval |
| Claude Code on Enterprise plan | Yes, if ZDR agreement covers Claude Code usage |
| Standard API (no enterprise agreement) | No |
The key requirement is that customers must contact Anthropic’s enterprise sales team, go through an approval process, and execute a separate ZDR agreement — it is not a checkbox available in the Anthropic Console.
Claude Code is a specific case worth noting: Claude Code can benefit from ZDR when deployed under a Claude Enterprise plan with an active ZDR agreement that explicitly covers Claude Code usage. This should be confirmed in writing during the contracting process.
To review the Anthropic DPA that applies to enterprise API customers, see our Anthropic DPA guide.
What ZDR Covers — And What It Doesn’t
ZDR eliminates post-response retention of standard API inputs and outputs. However, Anthropic has confirmed two categories that are always retained even under a ZDR agreement:
-
User Safety classifier results: Outputs from Anthropic’s safety review systems that flag potentially harmful content are always retained. This applies regardless of ZDR status and cannot be waived.
-
Flagged policy violations: If an input or output triggers a policy violation, the relevant content may be retained for up to two years for trust-and-safety and legal compliance purposes.
These exceptions exist because Anthropic must maintain records sufficient to investigate safety incidents and enforce its usage policies. They apply universally and are not subject to negotiation in a standard ZDR agreement.
For most enterprise use cases, these exceptions will not affect the practical benefit of ZDR: standard business documents, contracts, customer data, and code submitted via the API are not expected to trigger safety classifiers or policy violations.
What ZDR does not address:
- Geographic location of processing — inputs are still processed in Anthropic’s US infrastructure unless you route via Bedrock or Vertex AI
- Subprocessor access during active processing — ZDR governs storage, not the inference pipeline itself
- Training data use — ZDR prevents retention, but companies should separately confirm training opt-out terms in the DPA
ZDR vs EU Data Residency: Key Differences
These two concepts are frequently confused in enterprise procurement. They address different dimensions of data protection:
| Criterion | Zero Data Retention (ZDR) | EU Data Residency |
|---|---|---|
| What it governs | Whether data is stored post-response | Where data is processed geographically |
| Available through Anthropic API? | Yes, with enterprise agreement | No — requires AWS Bedrock or Vertex AI EU |
| GDPR principle addressed | Data minimisation (Art. 5(1)(e)) | Third-country transfer restrictions (Chapter V) |
| Requires separate agreement? | Yes — separate ZDR agreement | N/A — requires separate cloud provider setup |
| Default setting? | No | No |
| Can be combined? | Yes | Yes — EU hosting + ZDR agreement |
The key insight: ZDR and EU data residency are complementary, not alternatives. An enterprise team that requires both data minimisation (ZDR) and EU-only processing (residency) must address both independently:
- ZDR: obtained through the Anthropic enterprise sales process
- EU data residency: achieved by deploying Claude via AWS Bedrock Frankfurt (eu-central-1) or Google Cloud Vertex AI EU regions
For a full breakdown of EU data residency options, see our Claude EU hosting guide.
Why ZDR Matters for GDPR Compliance
Article 5(1)(e) GDPR — the data minimisation and storage limitation principle — requires that personal data is kept in a form which permits identification of data subjects for no longer than is necessary for the purposes for which the personal data are processed.
When companies use the Claude API to process personal data — customer records, employee documents, legal files — standard retention terms mean that data is held by Anthropic beyond the immediate purpose of generating a response. This creates retention obligations that must be documented, justified, and eventually fulfilled through deletion procedures under Article 17 GDPR.
ZDR removes that burden: if the data is not retained, there is no retention period to justify, no deletion schedule to enforce, and no ongoing storage exposure to manage. For organisations processing high volumes of personal data through the API, ZDR substantially simplifies GDPR compliance operations.
ZDR is particularly relevant for:
- Law firms and legal teams processing client documents via the Claude API
- Financial services companies inputting customer or transaction data
- Healthcare organisations processing patient-related data (Article 9 GDPR)
- HR teams using Claude to process employee data
- Any organisation with B2B contracts that include data minimisation commitments
For a broader review of GDPR obligations when deploying Claude, see our Claude Enterprise GDPR guide.
How to Request Claude ZDR
ZDR is obtained through Anthropic’s enterprise sales process, not self-served through the Console. The steps:
- Contact Anthropic sales: Initiate the enterprise discussion and indicate that ZDR is a requirement for your deployment.
- Receive approval: Anthropic reviews eligibility and approves ZDR for qualifying enterprise customers.
- Execute the ZDR agreement: A separate ZDR agreement is signed alongside your commercial agreement and DPA.
- Confirm scope: Verify in writing that ZDR covers your specific deployment, including Claude Code if applicable.
- Document internally: Record the ZDR agreement status in your Records of Processing Activities (RoPA) under Article 30 GDPR.
ZDR is not retroactively applied to data processed before the agreement was executed.
Legal Assessment: Is ZDR Sufficient for Your Use Case?
ZDR is a meaningful data protection control — but it is not a complete compliance solution on its own. Companies deploying Claude via the API under GDPR still need:
- A valid DPA (Auftragsverarbeitungsvertrag): The Anthropic commercial DPA under Article 28 GDPR must be in place for any personal data processing, regardless of ZDR status. For a review, see our Anthropic DPA guide.
- SCCs or an adequate transfer mechanism: Anthropic processes data primarily in the United States. Even with ZDR, the transfer of personal data for processing creates a Chapter V GDPR obligation. Standard Contractual Clauses (SCCs) are the standard mechanism, typically supported by a Transfer Impact Assessment (TIA).
- EU data residency where required: If your organisation’s contracts or sector regulation require EU-only processing, ZDR alone does not satisfy that. A separate Bedrock or Vertex AI deployment is needed. See our Claude EU hosting guide.
- A legal basis for processing: GDPR Article 6 requires a valid legal basis for any personal data processing, regardless of retention practices.
- DPIA where triggered: Certain high-risk processing activities under Article 35 GDPR require a Data Protection Impact Assessment, regardless of ZDR status.
ZDR is best understood as one layer in a multi-layer compliance architecture, not a standalone compliance solution.
Compound Law advises companies in Germany and the DACH region on AI procurement compliance, GDPR reviews, and DPA negotiations. If you are evaluating a Claude Enterprise deployment and need a legal review of your ZDR and DPA setup, contact us.
FAQ
Does Claude Zero Data Retention replace the DPA?
No. ZDR and the DPA serve entirely different functions. The DPA (Datenschutzvereinbarung) is the Article 28 GDPR processor agreement that defines roles, obligations, and safeguards for any personal data Anthropic processes on your behalf. ZDR is a separate contractual commitment about storage — specifically that API inputs and outputs are not retained after the response. Both instruments are required if you process personal data via the Claude API. ZDR reduces the retention scope, but does not establish the legal framework for processing.
What is still retained under Claude ZDR?
Two categories are always retained: User Safety classifier results, which flag potentially harmful content, and content that triggers a confirmed policy violation, which may be kept for up to two years. Everything else — standard API inputs and outputs — is not stored after the response is returned.
Is Claude ZDR available for Claude Code?
Yes, but only when deployed under a Claude Enterprise plan with a ZDR agreement that explicitly covers Claude Code. Without an Enterprise plan and explicit ZDR coverage, Claude Code does not carry the zero-retention commitment. Confirm in writing during contracting that Claude Code usage falls within the ZDR agreement scope.
Can ZDR and EU data residency be combined?
Yes — they address different dimensions and are not mutually exclusive. ZDR governs whether data is stored post-response. EU data residency governs where processing occurs geographically. To address both, combine a ZDR agreement with Claude deployed via AWS Bedrock Frankfurt (eu-central-1) or Google Cloud Vertex AI EU regions. Neither includes the other by default.
Which Claude plans are eligible for ZDR?
ZDR is available only to eligible API enterprise customers with Anthropic’s explicit approval and a separate ZDR agreement. It is not available on claude.ai Free, Pro, or Team plans. The Claude Enterprise plan through claude.ai does not automatically include ZDR — a separate agreement must be obtained through the enterprise sales process.
How do I request Claude ZDR?
Contact Anthropic’s enterprise sales team and indicate that ZDR is a requirement. ZDR is not self-service — it requires Anthropic’s approval and a separate ZDR agreement executed alongside your commercial terms. Confirm during the process that the ZDR scope covers your full deployment, including Claude Code if applicable.
Does Claude ZDR satisfy GDPR data minimisation?
ZDR directly supports Article 5(1)(e) GDPR by preventing post-response storage of API inputs and outputs. This removes the obligation to maintain a retention schedule for that data and reduces ongoing storage exposure. However, ZDR alone does not constitute full GDPR compliance — a valid DPA, a legal basis for processing, SCCs for the US transfer, and the other Article 5 principles must all be addressed separately.
What is the Anthropic Zero Data Retention agreement?
The Anthropic ZDR agreement is a separate contractual document executed by eligible enterprise API customers alongside their commercial agreement. It formally commits Anthropic to not retaining API inputs and outputs after the response, subject to the safety classifier and policy violation exceptions. It is obtained through the enterprise sales process and is distinct from the standard Anthropic DPA.