Garden privacy policy

Updated November 12, 2026

This policy explains how Diffuse Technologies Inc. ("Garden", "we", "us") accesses, collects, uses, stores, shares and deletes information when you use Garden at withgarden.ai. Garden connects your work systems, builds business context from authorized evidence, and lets you run Agents for work you approve. This policy covers the hosted service and its Google integrations, including Gmail, Google Drive, Docs, Sheets, Slides, Google Merchant Center, Google Ads and Google Analytics. Contact ryan@withgarden.ai with privacy questions or requests. Organizations operating their own copies of Garden are responsible for their own processing.

Google data we access and why

When you connect a Google account, Google's consent screen identifies the permissions requested. The data available depends on those permissions, your Google account's access, your connection settings and the operation requested. Google permissions can be broader than the operations Garden implements. Connecting an account does not give every workspace member or Agent access to its data.

We also process the Google account identity needed to identify a connection, the granted permission scopes, access and refresh tokens, connection settings, and operation results. OAuth tokens let our servers make authorized requests; they are not your Google password.

Derived, aggregated and anonymized data

Authorized features may turn Google data into summaries, answers, reports, extracted text, business context documents, citations and other artifacts, including aggregated metrics. These can contain personal or confidential information even when they do not reproduce the original record. We treat them as Google-derived data. The use, sharing, AI training and human-access restrictions in this policy apply to raw Google data and to derived, aggregated and anonymized Google data. Aggregation or anonymization does not authorize advertising, sale, general model training or an unrelated use.

Other information we collect

We process account and workspace identifiers, contact details, membership and authentication information, prompts, uploaded material, connection configuration, and operational records needed to provide and secure Garden. Hosting services process network and request information, such as IP addresses, access times and errors. Other connected services provide the account information, selected content or metadata and action results needed for the operations you request. For example, Vercel connections process project and deployment IDs, URLs, states, timestamps and available source revisions. Provider responses may contain personal or confidential information.

How we use information

We use information to provide the connections, search, answers, business context, reports and actions you request or authorize; maintain your account; enforce access permissions; operate and secure the service; and respond to support and privacy requests. A connection read does not itself start a model call. Agents may retrieve and process data in the background when you authorize their work, including scheduled work. A connection does not by itself authorize every Agent: Agent access must be approved separately, and read and write access are separate.

Google API data and Limited Use

Garden's use and transfer to any other app of information received from Google APIs will adhere to the Google API Services User Data Policy, including the Limited Use requirements. Our use of raw or derived user data received from Google Workspace APIs, including aggregated or anonymized data, adheres to these requirements.

We use Google user data only to provide or improve the user-facing features you request and that are described in this policy. We do not sell Google user data or transfer it to data brokers or information resellers. We do not use Google Workspace data for advertising, personalized or interest-based advertising, retargeting, creditworthiness or lending decisions. Google Ads data is used for the authorized campaign-management and reporting features described above. We do not use data obtained through Google Workspace APIs to develop, improve or train generalized or non-personalized AI or machine-learning models, and we do not transfer that data to third parties for such training. These restrictions include raw, derived, aggregated and anonymized Google user data.

Transfers of Google user data are limited to what is necessary to provide or improve the appropriate user-facing features, with your consent where required; security purposes; compliance with applicable law; or a merger, acquisition or sale of assets after obtaining explicit prior consent where Google's policy requires it. A transfer does not remove the Limited Use restrictions.

We do not allow humans to read Google user data except with your affirmative agreement to view specific data; when necessary for security, such as investigating abuse; to comply with applicable law; or for internal operations when data has been aggregated and anonymized as permitted by Google's policy. A support request alone does not authorize unrestricted reading of your files or email.

AI processing and who receives data

When you invoke an AI feature or authorize an Agent that uses a model, relevant prompts, authorized retrieved content, attachments and tool results may be sent to OpenAI API services to generate an answer or carry out that task. These inputs can include Gmail content, Google documents, Merchant Center records, Ads or Analytics reports and configuration, and their derived summaries. We use this processing for inference on your authorized work, not general model training. We do not opt Google data into model-training or optional training-sharing programs. We do not send Google Workspace API data to AI services that use it to train generalized models. Customers configuring other providers must preserve these restrictions.

Cloudflare and Vercel process requests and host application services. PlanetScale provides the hosted PostgreSQL database used for account, connection and application records. Cloudflare also provides private object storage for encrypted execution traces. These infrastructure providers process the raw or derived information stored or transmitted through their services as needed to provide Garden. OpenAI processes the inputs and outputs of hosted AI features. Authentication and operational service providers receive the account or operational information needed for their functions; connecting Google does not authorize distributing mailbox, file or merchant content to them for unrelated purposes.

Authorized content, summaries, reports and action inputs may also be returned to the authorized caller or sent to the destination, connected service or recipient specified in your task, such as the recipient of an email you authorize Garden to send. Workspace membership alone does not grant access to all connected data. We do not share raw, derived, aggregated or anonymized Google user data with advertisers, data brokers or model-training services. Our providers may retain data for service delivery, security or legal obligations; we do not promise zero retention of AI requests or responses.

Storage and protection

Google access and refresh tokens are stored server-side, encrypted using AES-256-GCM. We use HTTPS for browser access and Google API requests. Authentication, current workspace and source authorization, configured connection boundaries and separate Agent grants control access to data and operations. Revoking a grant stops subsequent authorized use of that grant. These controls reduce risk but cannot guarantee absolute security.

We store account and workspace records, connection settings, encrypted credentials and operational records. Authorized tasks may also persist original content or extracts, conversation inputs and outputs, reports, business context documents, citations and artifacts. Not every read is permanently imported, but a task that saves or summarizes data can create a retained copy. Relevant copies may be processed by the infrastructure and AI providers described above.

To diagnose failures and verify Agent work, Garden retains execution traces that may include prompts, model inputs and outputs, tool requests and results, and images. These traces can contain original connected data or derived data. Trace payloads are encrypted and stored in private Cloudflare object storage, with indexes and pending exports in PostgreSQL; access is subject to current workspace authorization. General operational telemetry is separate and excludes prompts and source content.

Retention, disconnection and deletion

We retain account and connection records while needed to operate your account and connections. We retain saved content and derived outputs while needed for the feature or workspace that holds them, unless you delete them through an available control or request deletion. Operational and security records are retained as needed for service reliability, abuse prevention, resolving disputes and legal obligations. We do not currently apply one universal automatic deletion deadline to all hosted data.

You can disconnect a connection in Garden and revoke Google authorization in your Google Account's third-party connections settings. Garden revocation stops subsequent access through that connection; provider token revocation is attempted where supported. You can separately revoke an Agent's grants to stop that Agent's subsequent use. Disconnection or revocation does not automatically delete previously saved messages, document extracts, merchant reports, summaries, conversations, artifacts, operational records or backups, and it cannot recall an email already sent.

To request access, correction, export or deletion of your account information, raw Google data or derived, aggregated or anonymized Google data, email ryan@withgarden.ai. Identify your account or workspace and the data or connection concerned; do not send passwords or tokens. After verifying your identity and authority, we locate relevant active records, delete eligible source copies and derived data, and address copies held by our service providers as applicable. We explain any limits, including legal or security retention or another organization's rights. Backup copies follow a separate lifecycle and may remain until they expire or are overwritten. Copies independently exported by you or sent to a recipient are controlled by that recipient and must be addressed separately.

Your choices and policy changes

You can review Google's consent screen, decline access, manage connection settings and Agent grants, stop authorized tasks, disconnect providers and request deletion. Privacy rights depend on applicable law. We publish policy changes here with an updated date and provide notice and obtain additional consent where required before materially changing our use of Google user data. Review each provider's own policy for information it holds independently.