Michaël Trazzi comments on Worst-case thinking in AI alignment