We use OpenAI API at work, it fails when translating children stories, the reason is violence. Either the model safety is shit or the AI companies are pushed by

We use OpenAI API at work, it fails when translating children stories, the reason is violence. Either the model safety is shit or the AI companies are pushed by some extremists groups to censor shit that is acceptable for children in Europe (Romania). But the most bullshit is when you give it a safe prompt, the mdoel generates a response and the safety checker kicks in and blocks the response because it thinks the model was too naughty.