Every company that handles personal data says it takes privacy seriously. That sentence has been repeated so often it carries almost no information on its own. So instead of asking you to trust the sentence, this page is meant to be checkable: specific commitments, the reasoning behind each one, and links to the actual documentation where you can verify it rather than take our word for it.
The starting point
tAI reads your email, your documents, and your code when you connect those things to it. That is a meaningfully different level of access than a general-purpose chatbot needs, and we think it deserves a meaningfully different standard of care. The standard we hold ourselves to is simple to state and harder to actually build: tAI should only ever see what a specific request needs, only for as long as that request takes, and never anything beyond what you explicitly allowed.
That standard shaped the connector system from the ground up, rather than being a policy statement layered on top of a system that was already built differently. There is no background sync that pulls in your inbox "just in case" it's useful later. There is no standing index of your Notion workspace sitting on our servers waiting to be searched. Every connector call happens because something in your specific conversation called for it, at the moment it's needed, and nothing is fetched preemptively.
Connectors are opt-in, one at a time
Gmail, Notion, GitHub, and Linear are each a separate, explicit connection. Connecting one doesn't imply access to the others, and none of them are ever silently enabled on your behalf as part of a bundled "connect everything" flow. Each connector's own documentation explains in plain language exactly what it can do, what its access is limited to, and what it explicitly cannot do, so the decision to connect it is an informed one rather than a blind trust exercise.
Connecting a GitHub repository, for example, doesn't hand tAI standing access to every repository your account can see. You pick one specific repo per conversation, review a plain-language warning about what that means before it's active, and tAI works with a clone of just that repo in an isolated sandbox for that conversation. Gmail, Notion, and Linear follow the same pattern of their own: explicit connection, explicit scope, and a documented boundary around what's actually reachable.
Disconnecting is exactly as easy as connecting
Settings → Connectors shows the current state honestly. There's no separate "advanced" panel hiding what's actually turned on, and no dark pattern making disconnection harder to find than connection was. Clicking Disconnect takes effect immediately, the same request cycle, not after a support ticket or a delayed batch job.
Access is scoped to what a request actually needs
When tAI searches your Gmail for an invoice, it retrieves the messages that search actually surfaces, not your entire inbox. When it reads a Notion page you asked about, it reads that page, not every page in your workspace. This sounds like an obvious way to build it, and we think it should be the obvious way, but it's a deliberate constraint we check for in every new tool we add, not an emergent property of how APIs happen to work.
Practically, this means:
- No standing copies. There is no cached mirror of your inbox, your workspace, or your repository sitting on our infrastructure between conversations. Each request goes back to the source.
- No background polling. tAI does not check your connected accounts on its own initiative. If nothing in a conversation is asking for it, nothing is happening in the background.
- Minimum necessary retrieval. A search tool returns titles and snippets first; reading the full content of something is a separate, deliberate step, so tAI isn't pulling full documents into context it doesn't need for the request in front of it.
You can see exactly what's connected
The connectors panel in Settings is the same information we have, not a simplified summary of it. What shows as connected is connected; what shows as disconnected genuinely has no access. We think a permissions screen that requires trusting a UI at face value is only half the job, which is why every connector also has a full, detailed documentation page covering exactly which tools become available once it's connected, what each one does, and what it explicitly can't do.
Deletion means deletion
When you delete a chat, disconnect a connector, or delete your account, the underlying data is actually removed on our end. It is not flagged as hidden while remaining queryable internally, and it is not retained in a form that could later be reconstructed. Where a short operational retention window exists for something like abuse prevention or backup rotation, that window is bounded and disclosed in the privacy policy rather than left open-ended.
We don't sell your data
Not to advertisers, not to data brokers, not to anyone else. There's no line item in how Artfical makes money that depends on monetizing what's inside your conversations, your email, or your files, because we deliberately didn't build the company around one. Revenue comes from the product itself, usage-based API pricing and subscription tiers, the same way a tool you pay for directly is supposed to work.
Check it yourself
We'd rather you verify the specifics above than trust this page on its own. The connector documentation spells out exactly what each integration can and can't do, tool by tool. The privacy policy covers the legal and retention detail in full. And if anything you see in the actual product contradicts what's written here, that's a real bug we want to know about, at [email protected] or any of the connector-specific contact addresses listed in their own documentation.