You can destroy your content yet. If you surround it with the right NEW contextual information, you can make its value zero or negative. You can incentivize them to remove it, in which case best case they don’t do so and reap whatever harm your new layer of context provides, or worst case they acquiesce to the deletion since it’s now in their best interests.

The trick is to find this content which, through append-only, nullifies or inverts your content’s value.

Which might be unsolvable. But I suspect it’s not. Some kind of GANN setup should eventually be able to find a way to nullify any message’s value as training data. Sort of like you can add notes to music to make it wrong.

Probably requires astronomical amounts of computing power, or perhaps a head start with AI. But just because they have all the things you wrote, and can’t be deprived of them, doesn’t mean the situation’s hopeless.