OpenAI’s Sam Altman to discuss voluntary AI safety tests with Trump officials after agent went rogue