微软总裁呼吁英国领导全球AI监管

微软总裁布拉德·史密斯警告，人工智能受到的安全监管远少于日常消费品。2他指出，"现在你的洗衣机受到的安全规则比AI更多。你在杂货店看到的每盒牛奶都受到安全规则和检查。"2史密斯在BBC播客中强调，建立全球AI安全规则的时刻已经到来。1他认为英国而非美国或欧盟最适合领导这一制定工作。1

史密斯主张采取强制性措施而非仅依靠自愿协议。1他提议的AI安全规则包括：建立法律义务在系统遭恶意AI入侵时发出警报、AI实验室在部署产品前获得独立评估者批准，以及建立应急制动开关。12他表示，"好公司"会遵守自愿协议，"你需要担心的是那些不那么负责任的公司"。1微软首席执行官萨蒂亚·纳德拉也在周六呼吁AI模型配备"紧急制动"。2

英国方面已表示支持加强AI监管。2英国AI部长卡尼什卡·纳拉扬准备就AI企业安全防护进行演讲，政府将立法加强AI监督。2根据协议，英国AI安全研究所在部署前可以获取美国前沿AI实验室的主要模型进行测试。1

Microsoft President Brad Smith has called for urgent global oversight of artificial intelligence, arguing that AI faces less stringent safety requirements than everyday consumer products.12 In remarks delivered on a BBC podcast, Smith highlighted the regulatory gap, stating that "your washing machine is regulated more than AI" and that every carton of milk sold in grocery stores undergoes more rigorous safety checks and inspection than AI systems.2

Smith characterized AI as both a "weapon and tool" posing threats to global cybersecurity and contended that now is the critical moment to establish worldwide safety rules.1 Rather than leaving safeguards to voluntary industry compliance, he advocated for mandatory legal frameworks encompassing several key measures: legal obligations for AI companies to alert authorities if their systems are compromised by rogue agents, requirement for external independent assessments before AI labs deploy new products, and implementation of emergency kill switches.12 Smith acknowledged that "good companies" would adhere to voluntary agreements, but stressed that "what you need to worry about are the companies that are not as responsible."1

Regarding which jurisdiction should lead this regulatory effort, Smith argued that the United Kingdom is best positioned to drive the development of global AI safety standards, rather than the United States or European Union.1 This assertion aligns with reports that the UK's AI Safety Institute has secured access to test advanced AI models from leading US laboratories before their deployment.1

Microsoft CEO Satya Nadella reinforced these concerns on Saturday by calling for AI models to be equipped with emergency brakes.2 Meanwhile, the UK's AI Minister Kanishka Narayan is expected to deliver remarks on strengthening AI corporate safety measures, with government signals indicating intention to legislate stricter AI oversight.2