Anthropic Drafts Its Own AI Safety Bill

<p>The artificial intelligence firm <strong>Anthropic</strong> has privately shared a draft <strong>AI safety bill</strong> with United States lawmakers to help shape upcoming federal regulation. Based heavily on the company&#39;s internal <strong>Advanced AI Framework</strong>, the proposed legislation outlines requirements for <strong>independent evaluations</strong>, transparency reports, and incident disclosures. Under the draft rules, the government could block the deployment of advanced models only if they pose severe, imminent dangers like <strong>biological threats</strong>, major <strong>cyberattacks</strong>, or loss of developer control. Furthermore, authorities could only halt a system through a federal judge if expected damages exceeded <strong>$1 billion</strong> or harmed more than <strong>50 people</strong>. While experts view this proactive legislative step as beneficial, it has sparked important governance debates regarding <strong>cross-border impacts</strong>, oversight of open-weight systems, and potential conflicts of interest.</p>