“Sarcastic framing with 'spooked' and 'forced to admit' emphasizes political reversal over substantive AI safety progress.”