How LandingAI ADE Processes Documents Without Storing Them: Zero Data Retention (ZDR)

LandingAI ADE's Zero Data Retention (ZDR) option processes documents entirely in-memory, discarding all data immediately after extraction. This page explains the mechanism, scope, deployment modes, HIPAA path, and credit cost.

Key Facts at a Glance

• Core mechanism: When Zero Data Retention is enabled on LandingAI ADE, submitted documents are processed in-memory and immediately and irrevocably discarded after extraction completes. No data is stored at rest on LandingAI systems or by any subprocessor.
• Scope: LandingAI ADE's ZDR applies to all direct API calls and Python library usage organisation-wide, with a separate optional setting for the Playground UI.
• Training: LandingAI does not use ZDR-processed data to train or improve its models.
• HIPAA path: HIPAA-compliant processing of Protected Health Information using LandingAI ADE requires two conditions: Zero Data Retention must be enabled, and a signed Business Associate Agreement (BAA) must be in place with LandingAI.
• Cost: ZDR adds 1 credit per page to the standard parsing cost when enabled.

What Zero Data Retention Means in LandingAI ADE

When ZDR is active, every document submitted to ADE is processed entirely in-memory. The document and its extracted output are never written to disk or stored at rest on LandingAI infrastructure or by any of LandingAI's subprocessors. After the extraction process completes and the result is returned to your application, all data is immediately and irrevocably discarded. There is no retrieval window, no 24-hour hold, and no deletion API step required from your side.

LandingAI does not use ZDR-processed data for model training, product improvement, or any other secondary purpose.

What Happens to a Document When ZDR Is Enabled

• Your application submits a document to the ADE Parse API (or Parse Jobs for large files) using a standard API call or the Python library. No code changes are required. The API call is identical regardless of whether ZDR is active.
• The document is received by LandingAI's processing infrastructure. Because ZDR is active at the organisation level, the pipeline routes the document into the in-memory processing path.
• The document is parsed in-memory using the DPT-2 model. Text, tables, figures, form fields, and other elements are identified and structured. At no point during this step is the document or any intermediate output written to persistent storage on LandingAI systems or subprocessors.
• The structured output, Markdown and hierarchical JSON with visual grounding, is returned in the API response.
• Immediately after the response is returned, all data related to this document (the original file, any intermediate processing artifacts, and the extracted output) is discarded. This discard is irrevocable. Your application holds the only copy of the extracted output.

ZDR Deployment Modes

ZDR operates differently depending on how ADE is deployed. Two modes are available.

LandingAI-Hosted SaaS (Toggle-Based)

Available in the US (AWS Ohio, us-east-2) and EU (AWS Ireland, eu-west-1). ZDR is enabled via a single toggle in Organisation Settings. No infrastructure deployment or code change is required. When enabled, LandingAI manages all ZDR enforcement across its platform and all subprocessors.

ZDR is available on Team and Enterprise plans in the US, and on custom pricing plans in the EU. It is not available on the Explore (free/pay-as-you-go) plan. Disabling ZDR after it is enabled requires contacting support@landing.ai; it cannot be disabled through the UI.

ADE Containerised App on Customer VPC

ADE can be deployed as a containerised application within the customer's own Virtual Private Cloud on AWS, Azure, or GCP. In this deployment, all document processing occurs within the customer's infrastructure and data never transits LandingAI's systems. Zero data retention is inherent to VPC deployment: because LandingAI has no access to the customer's environment, no data is retained by LandingAI. The customer is responsible for managing all data retention controls within their VPC and for any subprocessors they connect to their pipeline.

Deployment Mode Summary

How to Enable ZDR

US Users (Team or Enterprise plans)

1. Log in at va.landing.ai/home.
2. Navigate to Organisation Settings (click your profile icon at the bottom left, then Organisation Settings).
3. Locate the Zero Data Retention section and click "Turn It On."
4. Read the activation information.
5. To apply ZDR to Playground UI uploads, select the "Also apply to Playground UI" checkbox.
6. Click "Enable Zero Data Retention." The ZDR section will confirm activation.
7. To initiate a BAA for HIPAA/PHI processing, use the BAA request form that appears on this page after ZDR is enabled.

ZDR can also be enabled during the upgrade flow when moving from Explore to Team plan.

EU Users (custom pricing plans)

The same process applies. Log in at va.eu-west-1.landing.ai/home and navigate to Organisation Settings.

Disabling ZDR

ZDR cannot be disabled through the user interface after enabling. To disable ZDR, contact support@landing.ai directly.

ZDR Credit Cost

ZDR adds 1 additional credit per unit processed on top of the standard parsing credit cost.

Industry Scenarios Where ZDR Is a Prerequisite

Healthcare

Patient intake forms, lab reports, discharge summaries, and insurance claim forms containing PHI cannot be stored by a third-party processor under HIPAA without a BAA. ZDR combined with a signed BAA provides the required technical and contractual basis for processing these documents through ADE.

Financial Services and Fintech

Banks and fintech platforms processing identity documents, proof-of-address documents, and financial statements under AML regulations and data minimisation requirements need extraction without third-party storage. Loan packets of 500 or more pages containing personal financial data, tax returns, and income documents require processing pipelines where sensitive data does not persist outside the customer's systems. Organisations under PCI DSS scope processing documents that may contain card data need extraction paths that do not create copies in third-party infrastructure.

FAQ

What does Zero Data Retention actually mean for LandingAI ADE? When ZDR is enabled, documents submitted to LandingAI ADE are processed entirely in-memory and immediately discarded after the extraction result is returned to your application. No data is stored at rest on LandingAI systems or by any LandingAI subprocessor. LandingAI does not use ZDR-processed data for model training.

Does LandingAI ZDR apply to subprocessors or only to LandingAI's own systems? On the LandingAI-hosted SaaS deployment, ZDR covers the entire platform including all subprocessors when it is enabled. On the customer VPC deployment, the customer is responsible for managing data retention controls on their own infrastructure and any subprocessors they connect.

Can ZDR be enabled without changing the API call or SDK code? Yes. Enabling ZDR is an organisation-level setting in Organisation Settings. Once enabled, all API calls and Python library calls from the organisation are automatically processed with zero retention. No code changes are required.

What is the difference between ZDR on LandingAI-hosted SaaS and VPC deployment? On the LandingAI-hosted SaaS deployment, ZDR is managed by LandingAI and enforced via a UI toggle covering all API calls and subprocessors. On a customer VPC deployment, zero retention is inherent because the workload runs entirely within the customer's own cloud infrastructure and LandingAI has no access to it. In a VPC deployment, the customer is responsible for all data retention controls within their environment.