Privacy guide
Instinct AI privacy: training, permissions, and deletion
A practical explanation of Instinct's published data controls and the settings to review before connecting personal accounts.
Key takeaways
- Review the policy by information type and by the route used to supply it.
- Training settings, account disconnection, and deletion answer different questions.
- A published policy is evidence of stated practices, not an independent security audit.
- Check your current account controls before sharing information or delegating actions.
Does Instinct use your information to train AI?
Instinct permits some information to be used for model training. Its published rules distinguish general information, Google Workspace API data, and Vault materials. The table below explains those distinctions and the available controls.
This guide explains the published distinctions, not the security of an account we tested. Both official policies were revised August 26, 2026; we checked them September 20. Read the current versions alongside the controls offered during your own onboarding, especially if a particular data restriction determines whether you can use the service.
The controls solve different problems
Start by naming the outcome you want. Preventing a future connection, removing previously collected information, and changing training use are separate goals. A single successful settings change should not be taken as evidence that all three happened.
| Your goal | Published control or distinction |
|---|---|
| Change future training use | Opt out at app.instinct.com/settings; the policy retains a safety-review exception. |
| Understand Google API data | The policy excludes Google Workspace API information from model training and improvement. |
| Understand Vault materials | The terms restrict their use to providing your service and exclude model training. |
| Stop a connection | Disconnecting does not automatically delete indexed information. |
| Remove indexed external information | Review deletion controls at app.instinct.com/workspace. |
What to check when changing training settings
The opt-out does not undo previous training or stop every kind of service improvement. The policy also preserves the use of previously trained models. That is why changing a training preference is different from a request to remove stored information.
Record the date of your change, read the setting's explanation, and confirm that you are in the right account. Establish the scope of the choice before relying on it across accounts or devices.
Do not assume that the Google API exception also covers text you manually copy from an email into a chat. The route by which information reaches a service matters. When the distinction affects a sensitive task, ask the provider how that route is handled and retain its answer with your setup notes.
Disconnecting Google and deleting data are separate steps
Instinct's terms say indexed connected-service input may still be used after disconnection unless you follow the deletion process. The Workspace is the documented destination to review those controls.
Choose the desired endpoint: stop one connection, remove information from a completed task, or close the account. Use the option that explicitly describes your goal. Verify connection removal and deletion separately, retaining any confirmation. If the scope is unclear, describe the account and data category to support without attaching unnecessary copies.
Data privacy and permission to act need separate decisions
Instinct's terms allow actions through connected services, including transactions, and caution that actions may be irreversible. A decision about storing information does not settle whether an assistant may send a message, book a service, or spend money on your behalf.
Write down those action boundaries when you choose a task. For an appointment, identify whether the assistant may only research availability, may contact a provider, or may accept a time. For a purchase, identify the merchant, maximum total, and the point at which you want to review the details.
Check the actual settings that support those instructions. A conversational request for approval is useful, but it should not be treated as proof of a technical permission restriction. Verify the completed action in the relevant service before assuming that a confirmation message represents the correct result.
A useful review before sharing personal information
Start with the minimum context for one ordinary task. Check which files and accounts are involved, whose information appears in them, and where the result will go. Note your training preference, action boundary, and deletion route; revisit them when the task ends or policies change.
These checks help you make a decision based on your requirements. They do not establish that Instinct is safe for every purpose or that a policy statement has been independently verified. For a question the published documentation leaves open, use the official privacy contact at privacy@instinct.com.
Directory picks
Agents mentioned in this guide
Source notes
Primary references
We use first-party product pages and technical documentation wherever possible. Listing access can change, so check the linked source before relying on a number or feature.
Related resources