How your information is used

This notice covers the GAIAI Agents console, agent tasks, connected channels and enquiries made through this site. GAIAI is a brand. Contact [email protected] about privacy matters.

We use account and connection information to operate the service. You define the work your agent does in customer conversations and the business information it uses. When using your customers’ data, you are responsible for informing them about that use and having the necessary authority.

The information processed depends on the channel you connect, the permissions you grant and the work you request. Data comes from you, conversation participants and connected services.

• Account and connection: profile, email, account/organization/agent identifiers, permissions, OAuth credentials and channel session information.
• Conversation: messages, sender/recipient and conversation identifiers, timestamps, relevant attachments and replies. WhatsApp numbers and Instagram account details may be processed depending on the channel.
• Business information: profile, products, prices, FAQs, policies and information you share in CRM and record tables.
• Workspace: permitted emails, files, document/cell/presentation content, calendars, tasks, contacts and inputs/results of other tools used.
• Operations: task/delivery status, errors and audit records, session context and information shared in support requests.

Account details support sign-in and ownership checks; connection details provide access to an authorized account. Task and conversation content is processed to prepare replies and carry out requested work. Action and error records help track delivery, investigate problems and protect security.

Creating an account with Google, granting tool access and sending data to a model service are separate actions. Permission to connect an account is not blanket consent for all data processing or model training.

A task in a Google Chat Space, relevant conversation and tool results may be sent to the agent. Google Workspace access can cover email, Drive files, documents, spreadsheets, calendars and other permitted services. The requested Google permissions are not limited to files created by the application; available agent tools are also controlled by application permissions.

WhatsApp QR pairing establishes a WhatsApp Web session through open-wa. This is not the official WhatsApp Business Cloud API connection. Messages and information needed for the conversation are processed through that connection. Instagram uses Meta’s account sign-in and messaging services.

You prepare the agent’s task and business information in its management Space. Customer replies are sent to the connected WhatsApp or Instagram conversation. Depending on its task and tool permissions, the agent can read connected documents or update records.

Content sent to a model service can include task instructions, relevant conversation, information from files and tool results. It is sent from GAIAI’s execution environment to the selected model service.

Muse Spark 1.3 Standard, Muse, and Muse13 are connected to Meta Model API; Flash and Pro are connected to DeepSeek Open Platform. These options use an execution tool called Claude CLI; its name does not mean the data is sent to Anthropic.

Google Antigravity (AGY) models are temporarily hidden from new agent selections. Existing AGY model bindings remain runnable. Training use follows the service distinctions below.

Muse and Muse13 in GAIAI use Meta’s Contributor options. Under Meta’s official terms, these options are offered at a discount in exchange for permission to use model inputs and replies to train future Meta models. Conversation, document content and tool results included in the input can fall within this scope.

The Muse Spark 1.3 Standard option does not use inputs and replies to train Meta models. That distinction does not change the terms of GAIAI's Contributor options. A no-training condition does not mean data is never stored or transferred to another country.

DeepSeek’s privacy policy describes the use of data, including user inputs, for service and model development, the right to opt out of training use, and processing and storage in China. The applicable service agreement and data-use preferences determine the scope of this processing.

Google Antigravity’s individual-service terms include recording interactions, using them for product and machine-learning development and allowing staff review; data-use preferences can be changed in its settings. Enterprise access is governed by the relevant enterprise agreement.

Anthropic’s own Claude Code service has different terms for individual and commercial accounts. Commercial services do not use data for training by default, with exceptions such as explicit participation and feedback. Individual accounts have data-use preferences. These terms do not automatically apply to Claude CLI traffic connected to another model service.

Google Workspace data is subject to Google’s Limited Use rules. These limit use and transfer to what is needed for the feature provided to the user, and prohibit using or transferring the data to train or improve general-purpose AI models.

Consent to a connection or data transfer does not remove these limits. These rules apply to information obtained from Workspace separately from the model-service terms described above.

Your agent profile and working rules are saved and added to its instructions when it runs. WhatsApp delivery records contain message text and conversation details. Instagram event records contain message and sender identifiers and processing status. Connection details support account access, while task records track progress.

Deleting an agent closes the channel connection and deactivates the related product record; it does not erase all message delivery and audit records in the same action. Cleaning local execution sessions also does not erase database records, backups or copies held by external services.

Files in your Google account and messages sent on platforms remain in those services. Model-service retention terms are separate from training use. DeepSeek’s policy describes processing and storage in China; Google and Meta services may also involve international processing.

Revoke Google permissions in your account’s connections page. Manage WhatsApp linked devices and Instagram app access on those platforms. Removing access does not erase previously sent messages or created files.

Send requests for access, correction, deletion or objection to processing to [email protected] with the subject “GAIAI Agents data request”. Identify the account, channel and request. We may ask for additional information needed to verify ownership; do not send passwords or access keys.

The response to your request explains actions taken, any grounds for retaining records and any steps required in the relevant external service. Your statutory rights and your right to contact the competent data protection authority remain unaffected.

Your full name, business or team information and intended use entered in the site’s enquiry window, together with the page where you started the enquiry, are used to create a draft in your chosen WhatsApp or email application. Content and link information may be passed to that application when the draft opens; you send the message yourself. This choice does not authorize agent execution or model training.

We use account, event and contact details in support messages to investigate and answer your request. Share the information needed to explain the issue rather than an entire customer conversation.

We reflect changes to data processing or connected services in this notice. The date at the top shows the latest text update. Send questions about the notice to [email protected].