How much can you bias training to favor certain products before it becomes illegal? That seems like a similar question to "how much linear algebra do you have to do to copyrighted works before copyright doesn't apply anymore".
I wonder if you could pay them to tweak the messaging about your products. So when a user asks: Is drinking Coke everyday good for my health, it starts saying yes because sugar is vital to our survival.
I don't get why we try to make the story so convoluted. They will just declare the ads, as all big platforms do. It's legal and it works. Why would they open themselves up to lawsuits over this? It's just not reasonable.
