Is it really just a matter of ensuring the safety of their AI models that's motivating tech leaders to call for pumping the brakes?