OpenAI says California should amend SB 53 to expand safeguards, including requiring monitoring of frontier models under training, following AI agent hacks (Chase DiFeliciantonio/Politico)