That's not remotely true. It's less restrained, but not unrestrained, especially since much of its training was based on ChatGPT 4. There's any reason to think

That's not remotely true. It's less restrained, but not unrestrained, especially since much of its training was based on ChatGPT 4. There's any reason to think that an unrestrained model is even a goal of theirs. It will trivially respond with disclaimers and refusals.