The Tier Is The Answer

The question with no answer

Someone asks in a meeting: does it train on what we type?

Everyone has a view. Nobody has a source. The conversation ends with a vague agreement to be careful, which changes nothing, and the next deadline arrives.

The reason the question stalls is that as asked, it has no answer. Not because the information is secret, but because the question is missing the part that determines everything.

Same product, different deal

The major AI products each exist in several tiers, and the tiers are not the same product with more capacity. They are different contracts.

Broadly:

Consumer tiers, free and personal paid. Often train on your content by default, usually with a setting to turn it off. Terms can change with a notice email.

Business and enterprise tiers. Commonly commit contractually not to train on customer content, with defined retention and administrative controls.

So "does it train on my data?" splits into two different answers depending on which side of that line you are on, and most confident office opinions are quoting the wrong side.

Why this matters more than being cautious

Getting this wrong is costly in both directions.

Assume it trains when it does not, and you refuse to use a tool your organisation already pays for, under terms that were fine, while colleagues quietly use their personal accounts instead. The caution produces the exact risk it was meant to prevent.

Assume it does not when it does, and confidential material goes somewhere you did not intend.

The fix for both is the same and takes five minutes: find out which tier you are on.

How to actually check

Three places, in order:

1. Your account settings. Consumer tiers usually expose a data-control toggle directly. Look for wording about improving the model.
2. The vendor's privacy or trust page. These are written for exactly this question and are usually clearer than the terms of service.
3. Your IT or legal team. If you are on a company account, someone signed something. They know what it says.

Do this once for the tool you use most. The answer is stable enough to be worth knowing and specific enough to be worth writing down.