AI is a choice
Connecting a mailbox and enabling the assistant are separate actions. You can prepare a campaign without AI. Activation requires visible consent to processing the relevant exchanges and a reply mode you choose.
The context provided
The relevant incoming reply, limited recent history, messages already sent in the campaign and your approved instructions may be sent to Anthropic. The model is not given unrestricted access to your mailbox. Google data is not used to train general-purpose models or for advertising.
Review, sending and human handoff
A proposed reply in review mode waits for your approval. In automatic mode, replies follow the brief and rules you define. Unsupported requests should be handed to a person. The confidence score remains an estimate, and an AI reply can contain errors.
Retention and access withdrawal
Incoming text cached for campaign conversations or AI review is encrypted and retained for up to seven days under the current pilot policy. Disabling AI stops that processing mode; account deletion and Google revocation are separate actions. Read the privacy policy for full scope, backups and your rights.