Partial, narrow rules ahead: safety testing, transparency, liability. A full federal AI regime by 2027 seems unlikely.
