How China and the US can work together on AI security - FT中文网
登录×
电子邮件/用户名
密码
记住我
请输入邮箱和密码进行绑定操作:
请输入手机号码,通过短信验证(目前仅支持中国大陆地区的手机号)。如果该手机号尚未注册,将自动创建 FT中文网账号。短信可能需要几分钟送达,验证码15分钟内有效,请耐心等待:
请您阅读我们的用户注册协议和隐私权保护政策,点击下方按钮即视为您接受。
人工智能

How China and the US can work together on AI security

Xi-Trump meeting should build on plans for dialogue on shared risks
00:00

{"text":[[{"start":7.76,"text":"It may turn out that the most significant development linked to this week’s summit between President Donald Trump and Xi Jinping happened days before the Chinese leader’s scheduled arrival in Washington on Wednesday night. US Treasury secretary Scott Bessent and China’s vice-premier He Lifeng last weekend agreed to establish a formal US-China dialogue on AI. The US has also proposed a mutual notification mechanism about AI incidents serious enough to raise national security concerns — reminiscent of the nuclear hotline set up between Washington and Moscow after the Cuban missile crisis. Together these could mark first steps, however tentative, towards applying some guardrails to the world’s most perilous technological race."}],[{"start":53.56,"text":"Recent apocalyptic warnings have focused on the risk of AI turning against its human creators. But the fierce competition between the US and China for AI leadership means that agreeing on ways to slow the development of cutting-edge models — however desirable — is likely to be the most contentious issue between Washington and Beijing. Trump has dismissed AI doomsday scenarios as a “hoax”. Beijing sees any US talk of curbing frontier models as an attempt to lock China into second place in the AI race — although since Chinese open-source models have now become a major export, it is in Beijing’s interest to reassure over AI safety. (US officials suggest China seems less concerned about AI’s risks to humanity than about its risks to Chinese Communist Party control)."}],[{"start":100.76,"text":"Yet for all the suspicion, Washington and Beijing have shared vulnerabilities on which it would be in their interests to co-operate and which could form the basis of the AI dialogue. The first is the risk of highly capable AI falling into the hands of non-state actors. Advanced AI is making it potentially much easier for terrorists, rogue non-state groups or even organised crime to engage in activities ranging from developing bioweapons to launching cyber attacks on telecoms, energy or transport networks or financial markets."}],[{"start":131.36,"text":"An associated danger is the risk of AI agents escaping testing environments and themselves conducting hacks of foreign companies or infrastructure. Agreeing baseline standards for controlling access, identifying customers, testing models and establishing containment mechanisms could begin to reduce risks. The speed with which an AI-directed incident could unfold, compressing the time available to work out who or what was responsible, also risks triggering an escalatory spiral of retaliation. That makes the US AI incident “hotline” idea one that would make sense for Xi to endorse in his meeting with Trump."}],[{"start":168.08,"text":"Clear rules of engagement are similarly needed on AI in weaponry and nuclear arms systems. Here, an important foundation already exists: Xi agreed with US President Joe Biden in November 2024 that humans should retain control over nuclear weapons use. But no progress has been made since then, even as AI technology has made giant leaps. Security experts from both sides have since 2019 been developing recommendations that should feed into the AI dialogue. They include “absolute” red lines prohibiting AI systems from making autonomous decisions to launch nuclear weapons; mandatory human authorisation for any AI-driven cyber attack on an adversary’s nuclear control network; and agreeing precise definitions of what counts as “meaningful human control” over weapons systems."}],[{"start":217.1,"text":"The first US-Soviet nuclear hotline was a response to a stand-off that pushed the world to the brink of conflict. The task for US and Chinese AI negotiators is to put a safety framework in place before any such near miss — or actual cataclysmic event — caused by a technology that is developing at breakneck speed."}],[{"start":239.98,"text":""}]],"url":"https://audio.ftcn.net.cn/album/a_1790154652_7614.mp3"}

版权声明:本文版权归FT中文网所有,未经允许任何单位或个人不得转载,复制或以任何其他方式使用本文全部或部分,侵权必究。
设置字号×
最小
较小
默认
较大
最大
分享×