Sources: the White House and Anthropic are working on a framework that would assess the severity of AI security flaws, a sign that negotiations are progressing (Politico)