A bipartisan, first-in-the-nation bill would require major AI developers to disclose risks, report safety incidents and submit to annual independent audits.