1) More and more AI content will end up on the internet that even the AI won't be able to distinguish from human-generated. It will be like the snake eating its own tail.
2) Bot armies filling the internet with wrong info to tweak the AI into believing falsities.
3) Human intermediaries between AI interfaces and the AI so that certain questions can't be asked and certain answers won't be provided.
You are correct.
One thing that might happen is rival AI vendors escalating a war of BS attempting to poison, and thereby discredit, their rivals’ AIs.