The companies are obviously afraid of a journalist showing "I asked it for X and got back this horrible/evil/racist answer" - the "AI Safety" experts are capitalizing on that, and everyone else is annoyed that the tool gets more and more crippled.