Comment by comp_throw7

They don't currently claim to confidently believe that existing models are sentient.

(Also, they did in fact give it the ability to terminate conversations...?)