People seem to be continuously outraged by these AI subscriptions banning third party use. However, the usage patterns of the intended apps likely differ hugely from those of the third party ones.

For example, basically every first party agent harness aggressively caches the input tokens to optimise inference, something that third party harnesses often disgregard, or are fundamentally incompatible with as they switch agents for subtasks and the like.

To extend this use case though, how much do poeple expect to be able to use the internal API's of the apps they subscribe to?

If I buy an Uber One subscription, am I then justified reverse engineering the gazeteer API from the app and reusing it in other apps I use? What about the speech to text API MS Teams must use for transcribing meetings as part of a business standard subscription?

I think these are obvious and emphatic breaches that no reasonable person would expect to be justified in, maybe miffed if your clever hack gets banned, but being banned would be considered fair play.

I fail to see the distinction.