> I'd say OpenAI is rightly concerned that fine-tuning on chain of thought responses from this model would allow for quicker reproduction of their results.

Why? They're called "Open" AI after all ...