How does Gate.AI Zero Data Retention (ZDR) work?
Zero Data Retention (ZDR) on Gate.AI enables privacy-forward data handling by restricting the retention period of API request and response content at the routing layer and, by default, not using customer prompts for product improvement programs.
Teams that carry production-grade LLM traffic through the gateway need to distinguish what each layer retains: what the intermediate layer retains versus what upstream model providers retain. Gate.AI treats ZDR as the default privacy posture on the control plane. As of June 2026, the pricing page shows that Enterprise-tier customers can obtain contract-level Enterprise ZDR and DPA terms.
The following explains what ZDR means in Gate.AI’s architecture, how data is handled during routing calls, the differences between the default tier and Enterprise commitments, the relationship between console logs and retention strategies, and the key points teams still need to verify with upstream vendors—so you can understand it alongside the overall framework of Gate.AI Enterprise AI Data Privacy, together with RBAC and organizational controls.
What Is Zero Data Retention (ZDR) on Gate.AI, and Why Does It Matter?
Gate.AI’s Zero Data Retention (ZDR) is a data-handling approach. The platform limits retention of customer API traffic within Gate.AI’s control system and, by default, does not use the relevant content to improve Gate.AI products.
ZDR matters because LLM API calls often include prompts that may contain personal data, trade secrets, or regulatory information, along with generation results and metadata. Without clear routing-layer policies, teams struggle to answer audit questions: Are prompts stored? How long are they retained? Do they enter the provider’s training pipeline? Gate.AI addresses this routing-layer portion by documenting a default posture of not retaining data and not using it for product improvement, plus Enterprise ZDR and DPA options.
Gate.AI’s ZDR applies to data processed through Gate.AI as a routing gateway. Gate.AI cannot replace each upstream model provider’s own retention policies. After requests are forwarded, each vendor’s contract terms apply independently.
What Conditions Must Be Met Before ZDR Policy Takes Effect?
Before Gate.AI’s ZDR posture applies to a given request, the caller must authenticate using a valid organization API Key and send traffic to Gate.AI endpoints such as xx/openai/v1 or xx/anthropic.
The organization must already be onboarded in Gate.AI, and the credentials must be valid. Members create a Key via Console → Settings → API Keys. Integrations that have never gone through Gate.AI and instead connect directly to upstream providers fall outside the scope of Gate.AI ZDR.
Routing and guardrail settings influence which models the traffic reaches, but they do not replace the platform-level privacy baseline described on the pricing page. Enterprise customers who need contract-level ZDR typically sign an Enterprise agreement rather than relying only on self-serve tiers.
How Does ZDR Take Effect Step by Step as Data Flows Through Gate.AI?
After an application sends prompts to Gate.AI, the routing layer receives the request. It executes routing rules and organizational policies such as guardrails, then forwards compliant traffic to the selected upstream model.
During this process, Gate.AI processes request and response content to complete the call, enforce quotas, and generate billing and operational records. Official materials state the default behavior is not to use user data for product improvement, and it positions ZDR as the privacy commitment. Under the default policy, the routing layer acts as a short-lived control node rather than a long-term content archive.
After the upstream model returns a result, Gate.AI sends the response back to the client. Content may exist briefly in the processing chain required to deliver the service. The ZDR commitment concerns retention for training and product improvement purposes at the Gate.AI layer, not the elimination of all operational processing.
Figure 1. Gate.AI ZDR limits long-term retention at the routing layer and forwards to upstream; the Enterprise edition adds Enterprise ZDR and DPA contract terms (as of June 2026).
How Do Default-Tier Privacy and Enterprise ZDR Differ?
As of June 2026, the pricing comparison table shows that Free and Pay-as-you-go data privacy defaults to not retaining data and not using it for product improvement, and notes that options may be configured within a certain scope.
The Enterprise list adds Enterprise ZDR and DPA, indicating that organizations requiring formal commitments from the data processor can obtain written assurances. The Enterprise edition also includes organization and permissions management, SSO, and dedicated support—often paired with regulated deployments—but these are different dimensions from the ZDR contract language itself.
| Privacy Dimension | Free / Pay-as-you-go (as per pricing page) | Enterprise |
|---|---|---|
| Stated default posture | Default: no retention; no use for product improvement (configurable) | Enterprise ZDR + DPA |
| Contracting channel | Self-serve tiers | Enterprise sales / contract |
| Organizational RBAC | Lists organization-related capabilities under Enterprise edition | Included |
| Typical procurement questions | Is it acceptable for development and normal production? | Does it satisfy legal and procurement review? |
Teams typically upgrade from Pay-as-you-go to Enterprise due to legal requirements for DPA and named ZDR terms, not because the routing mechanism is completely different. DPA and Enterprise Data Compliance explains how Enterprise contracts incorporate the above commitments into regulated procurement workflows.
What Is the Relationship Between Gate.AI Console Logs and ZDR?
Gate.AI provides operational visibility in Console → Logs, with three tabs: Generation (single API requests), Tasks (asynchronous jobs), and Sessions (multi-turn aggregation).
Logs are used for troubleshooting, token and cost reconciliation, and usage analytics. They are not the same category as long-term datasets used for model training or product improvement. Privacy reviews should follow up on which fields appear, who can access them under RBAC, and how long operational records are retained. These questions complement ZDR policy and cannot substitute for one another.
Based on the role matrix, super admins and Tier-1 administrators have broader visibility into usage than typical members. As a result, log access intersects with access-control design, even if retention policies restrict training purposes.
What Common Misconceptions Exist About ZDR, and How Should Teams Verify?
The most common misconception is assuming Gate.AI ZDR automatically covers all upstream vendors. Gate.AI forwards content to external providers, and those providers maintain their own retention, logging, and training policies. A complete enterprise review must map approved models to vendor terms and record sub-processors.
Another misconception is equating ZDR with "zero logs." Operational records needed for billing and support may still exist for a reasonable period during platform operations. Teams should document what Gate.AI commits to at the routing layer and what the application side still must handle—client caching, prompt logs within application libraries, third-party observability tools, and so on.
Misconfigured API Keys or key sharing can cause traffic to fall outside the intended organizational boundaries, which is unrelated to ZDR. ZDR covers Gate.AI’s handling; credential hygiene is still critical.
Summary
Gate.AI Zero Data Retention (ZDR) limits retention of API traffic at the routing layer and defaults to not using prompts for product improvement. This commitment applies to traffic authenticated and forwarded via Gate.AI endpoints. As of June 2026, Free and Pay-as-you-go document the default no-retention posture, while the Enterprise edition provides Enterprise ZDR and DPA. Console logs serve troubleshooting and billing; they fall under a different category from retention policies for training purposes. Teams must still independently review each upstream model provider’s data terms, and include the approved models, sub-processors, and the access-control design in Gate.AI Enterprise AI Data Privacy in their compliance assessment.
FAQ
Q: What is Zero Data Retention on Gate.AI?
A: Per Gate.AI’s official pricing and product documentation, Gate.AI ZDR limits retention of API traffic within the routing layer and, by default, does not use customer prompts for product improvement.
Q: Does Gate.AI ZDR apply to data policies of upstream OpenAI or Anthropic?
A: No. Gate.AI ZDR covers only the routing layer within Gate.AI’s control scope. Each upstream provider applies its own retention and training terms after receiving the forwarded request.
Q: Which pricing tier includes Enterprise ZDR and DPA?
A: The Enterprise tier listed on gate.ai/pricing specifies Enterprise ZDR + DPA. Free and Pay-as-you-go describe the default posture of no retention and no use for product improvement, and note that configurable options are available.
Q: If ZDR applies, can admins still view prompts in Gate.AI logs?
A: Console logs provide operational records for API calls, tasks, and sessions. During compliance review, you should assess ZDR policy together with the log contents and RBAC visibility.
