Part of the problem is that these tech companies aren’t being required to back up their claims. They are somehow allowed to get away with minuscule disclaimer text that says “artificial intelligence models can make mistakes, check the output” and say complete bullshit about what it can do in interviews.