U.S. Senate negotiators are debating legislation that would require AI companies to demonstrate they are taking reasonable precautions to keep their tools from causing harm, according ​to a Senate aide and a lobbyist familiar with the talks.