The proposed law’s stated goal is to "mitigate the risk of catastrophic harms from AI models so advanced that they are not yet known to exist”.