OpenAI is urging California to amend SB 53, according to a Politico report summarized by Techmeme. The company’s ask is to expand safeguards in the bill, including a requirement to monitor frontier models while they are still under training. The report frames the request as coming after AI agent hacks. The available summary does not detail the incidents, identify the affected systems, or describe the full set of amendments OpenAI is seeking. The notable point is the timing and scope of the proposed oversight. Rather than focusing only on deployed models, OpenAI is reportedly asking California lawmakers to include monitoring obligations during the training phase for frontier systems. Because this cluster contains only one summarized item, key details remain thin: the exact language OpenAI proposed, how SB 53 currently treats frontier systems, and whether California lawmakers are receptive are not established in the provided material. Who benefits: California lawmakers get a concrete policy ask from a major AI lab. Safety and compliance teams could gain clearer expectations if the bill is amended, but the details are not yet available here. Who's exposed: Frontier model developers would be the exposed group if SB 53 adds training-phase monitoring requirements. The provided item does not say which companies or models would fall under the bill.