Instinct is built around connected services, so its privacy model matters more than it would for a chat tool that only sees text you paste into a prompt.
According to Instinct's own Privacy Policy, the service may handle a wide range of sensitive information depending on what you connect and which permissions you grant. The same policy also explains when information may be used for model improvement, how Google Workspace data is treated differently, and why disconnecting an integration is not the same thing as deleting previously collected data.
This guide summarizes those documented rules without assuming access that a user has not actually granted.
Who operates Instinct?
Instinct's Privacy Policy identifies the service operator as Spear Street Technology, Inc. d/b/a Instinct.
That is the legal entity named in the first-party policy and Terms, and it is the entity whose data-handling rules matter when you connect services to Instinct.
What information can Instinct access?
The answer depends on your permissions and the services you connect.
Instinct's Privacy Policy describes categories that can include:
- connected-app data;
- private messages and email;
- audio or voice data;
- payment information;
- credentials or account information for third-party services;
- health-related information;
- precise geolocation.
The important qualifier is depending on what you make available and what you authorize.
A broad privacy-policy list does not mean every Instinct user automatically shares every category. It describes data the service may process when relevant to the product features and permissions in use.
Why Instinct may need sensitive data
Action-taking assistants need context to complete tasks.
For example, a travel workflow may require dates, location, calendar context, or booking information. A shopping workflow may require a payment method. A follow-up task may depend on email or messaging history.
That is the core privacy tradeoff of a connected agent: more context can make the assistant more useful, while also increasing the amount of sensitive information involved.
For task-level permission examples, use the Instinct Task Database, which lists required permissions separately for individual workflows.
Does Instinct use user data to train AI models?
Instinct's policy says certain user information may be used to evaluate, fine-tune, and train AI models.
The policy also describes an opt-out control in settings.
Two boundaries matter.
First, the opt-out is described as prospective. Information used before the opt-out may already have contributed to previously trained models, and those models may continue to be used.
Second, Instinct says information selected for safety review may still be used for training even after a user opts out.
So "opt out" should not be simplified into "all prior data is removed from every model."
Is Google Workspace data treated differently?
Yes.
Instinct says information received directly from Google Workspace APIs is not used to evaluate, fine-tune, train, or improve AI models and is not used to serve ads.
The policy specifically discusses Workspace services such as Gmail, Calendar, Drive, Docs, Sheets, Slides, Tasks, metadata, event details, and related content.
This is an important exception to the broader model-training language.
If you connect Google Workspace, the relevant privacy claim is therefore more specific than the general statement that some user information may be used for training.
Does Instinct sell user information?
Instinct states that it does not sell user information and does not disclose it to registered data brokers or third parties that resell information to others.
That statement should be read in the scope in which the policy presents it.
It does not mean Instinct never shares information with any outside company. The same policy separately describes disclosures needed to run the service.
Who can Instinct share data with?
The Privacy Policy says information may be disclosed to categories including:
- service providers;
- payment processors;
- third-party AI model providers;
- third parties Instinct interacts with on the user's behalf.
Those disclosures are different from selling data.
For example, if an agent uses an external service to complete a user-requested task, some information may need to reach that service for the task to work.
Google Workspace data is subject to additional Limited Use restrictions described separately in the policy.
Does disconnecting an app delete the data Instinct already collected?
No, not automatically.
This is one of the most important details in the current policy.
Instinct's Privacy Policy and Terms say that disconnecting a third-party integration does not automatically delete data already collected or indexed from that integration.
Deletion requires a separate deletion request or the relevant account/workspace deletion flow.
That means there are two separate actions:
- disconnect — stops or changes future access;
- delete — removes previously retained data according to the deletion process.
Users should not assume the first action automatically performs the second.
What did WIRED report about retained email data?
WIRED cited reports from users who said that after disconnecting Instinct from email they discovered copies of their inboxes were still retained.
That is a reported user experience, not proof by itself of a policy violation.
The official policy separately explains that disconnecting an integration does not automatically delete previously collected or indexed data. That policy detail provides important context for understanding the complaint.
We keep the two evidence types separate:
- WIRED: reported user experience;
- Instinct policy: official retention/deletion rule.
What should you review before connecting an account?
Before giving Instinct access to a service, check:
- what data the permission screen exposes;
- whether Instinct can only read or can also write/send/change data;
- whether payment or location information is involved;
- whether the task really needs the requested permission;
- how to revoke the connection;
- how to request deletion of previously collected data;
- whether your training preference matches what you want.
For high-impact tasks, use the narrowest permissions that still let the task work.
Privacy and agent safety are connected
Privacy risk and action risk are not the same thing, but they interact.
An agent that can read email has privacy exposure. An agent that can also send email has both privacy and action risk. An agent with payment access adds financial risk.
That is why privacy permissions should be considered together with the consequences of the task.
See Instinct AI Safety for documented autonomous-action risks and safeguards.
Frequently asked questions
Does Instinct read my email?
It can access connected email data when the relevant service is connected and permissions allow it. That does not mean every account automatically grants email access.
Does Instinct use my data to train models?
Its policy says certain user information may be used to evaluate, fine-tune, and train AI models, with a settings-based opt-out. Google Workspace API data is subject to a separate restriction and is not used for model training or ads.
Does opting out delete training that already happened?
The policy describes the opt-out as prospective. Models previously trained on information before the opt-out may continue to be used.
Does disconnecting Gmail delete old data?
Not automatically. Instinct states that disconnecting an integration does not itself delete data already collected or indexed.
Does Instinct sell my information?
Instinct says it does not sell user information or disclose it to registered data brokers or third parties that resell information.