Connectors, permissions and the word no
A connector lets an AI read a live calendar, folder or mailbox instead of only what you paste in. That is powerful, and it is also the first level where the important lesson is about saying no.
In one family session, we connected an AI to a shared calendar so it could find the best free evening before a science fair. It found the evening in about four seconds - and then, unprompted, offered to also read the shared photo album "in case that helps too." It hadn't been given access to the album. It had simply asked, the way a helpful person might ask to borrow something they don't actually need yet.
That small moment is level 6 in miniature. A connector is what lets an AI read a live source instead of only what you paste in - genuinely useful, and also the first point in this pathway where the important skill is not granting access but withholding it.
What is an AI connector?
Up to this level, every answer came from something typed into the chat by hand. A connector changes that: it lets the AI read a live, ongoing source directly - a calendar, a shared folder, a mailbox - with permission a person granted and can take back at any time.
- Read-only by default - it can see the calendar, not edit it.
- Scoped, not total - access to one folder, not the whole drive.
- Reversible - the connection can be switched off as easily as it was switched on.
What is an AI connector for?
The upside is real: an AI that can check a live calendar and find three free evenings before Friday is doing something a pasted-in message from last week cannot do, because the calendar has changed since then. Connectors are what make an assistant useful for things that are actually happening now, not just things you already wrote down.
Access is a decision, not a default
The habit worth building here is treating every connection as a deliberate choice, made out loud, rather than something that happens because a screen offered it and nobody said no. "Should this be connected" deserves the same two-second pause as "should this door be unlocked."
One honest limit: a connector cannot tell the difference between a folder you meant to share and one you forgot was inside it. Scoping access narrowly - one calendar, one folder, not "everything" - is the child's job, not the tool's. The next article turns that sorting exercise into a concrete family rule about what should never be connected at all, and the one after that covers the single most important control on this whole level: taking access back.
Questions we get asked
What is an AI connector?
It is a permission that lets an AI read a live, ongoing source - such as a calendar, a shared folder or a mailbox - directly, instead of only working from text pasted into a chat. Good connectors are read-only by default, scoped to a specific source rather than everything at once, and can be switched off at any time.
Is it safe to connect an AI to a family calendar or photo folder?
It can be, with two conditions: the connection should be scoped narrowly to the specific source needed, not broad access to everything, and a parent should be part of deciding what gets connected rather than accepting a default. Treat each connection as a deliberate choice, not something that happens automatically.
What should I not connect an AI to?
As a starting rule, anything containing other people's private information, financial details, or content a family has not explicitly agreed to share should stay disconnected. The next article in this level covers this sorting exercise - fine to connect, connect with care, or never - in more detail.