Most people have that suspicion, myself included. But the API models stay verifiably same, and also OpenAI would have little reason to lie about not updating Ch

Most people have that suspicion, myself included. But the API models stay verifiably same, and also OpenAI would have little reason to lie about not updating ChatGPT. Otoh, I can easily imagine helper/censorship models and chat’s system prompt being updated. System prompt doesn’t change capabilities much though, and the censor model behaviour doesn’t change the output, just cuts it off when it discovers copyright or other violations (the case of chat not being able to recite the Dune poem, DALLe not being able to produce certain images etc)