There is AI in Tradelynx, and there is a rule about it that predates every feature: it is never the sole basis for a decision that goes against a person.
What that means concretely
A trader's credentials are approved by a person, never automatically. Where AI helps read a document, a human decides what happens because of it. Rejection, suspension, anything with a person on the receiving end has a person on the deciding end.
That is not a limitation we tolerate. It is the design. A model can be wrong in ways nobody predicted, and an appeal to "the system decided" is no answer to somebody whose livelihood it affected.
The second question: did the data need to go anywhere
Every AI feature is also a data-sharing decision, because the input goes to a vendor. So the inputs are minimised to what the model actually needs, the vendors are the two we have approved and no others, and features that would require sending more than necessary do not get built that way.
Some features are deliberately not AI at all. The maintenance forecasting on your passport is arithmetic on dates you can see, because it did not need to be anything cleverer, and arithmetic does not require sending your property details anywhere.
Why write this down
Because "we use AI" now appears on every product page in existence, and it tells you nothing. The questions that matter are who is accountable and where the data went. Those have answers here, and they are written into how the platform is governed rather than into a marketing page.