Do you use the chatgpt website or the api? I suspect these are problems related to the openai's interface itself rather than the models. I have problems getting chatgpt to find me things that it may think it may be illegal or whatever (even if they are not, eg books under CC license). With kagi assistant, with the same openai's models I have not had any such issues. I suspect that should hold in general for api calls.
Also, kagi's deepseek r1 answers the question about about propaganda spending that it is china based on stuff it found on the internet. Well I dont care what the right answer is in any case, what imo matters is that once something is out there open, it is hard to impossible to control for any company or government.
Well, I do, and I'm sure plenty of people that use LLMs care about getting answers that are mostly correct. I'd rather have censorship with no answer provided by the LLM than some state-approved answer, like O1 does in your case.
Oh wow, o1 really refuses to answer that, even though the answer that Deepseek gives is really tame (and legal in my jurisdiction): use software to record what's currently playing on your computer, then play stuff in the YTM app.
I ask O1 how to download a YouTube music playlist as a premium subscriber, and it tells me it can't help.
Deepseek has no problem.