Donald Trump signs watered-down AI vetting order after Maga infighting - 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。
商业快报

Donald Trump signs watered-down AI vetting order after Maga infighting

President’s directive will allow the US government to gain early access to cutting-edge models
00:00

{"text":[[{"start":7.7,"text":"Donald Trump has signed a long-awaited executive order that creates a “voluntary framework” for the US to gain early access to cutting-edge AI models, after officials raised concerns about the security risks they posed. "}],[{"start":22.05,"text":"The order is a watered-down version of an earlier draft, which was due to be signed last month but Trump pulled over concerns that federal oversight of frontier AI models might hamper the growth of leading labs including OpenAI, Google and Anthropic."}],[{"start":37.95,"text":"It comes after senior members of the administration, including Treasury secretary Scott Bessent, raised the alarm about the capabilities of Anthropic’s soon-to-be-released Mythos model, which is capable of identifying and exploiting cyber security vulnerabilities."}],[{"start":53.95,"text":"The US has been racing to give agencies access to Mythos before it is publicly available, while other governments and businesses around the world have been clamouring for access in order to patch security vulnerabilities that may be identified by the model."}],[{"start":67.85000000000001,"text":"Under the order signed on Tuesday, AI labs will be asked to submit their new models for security reviews “for a period of up to 30 days” before they are publicly released. The original executive order — a draft of which was seen by the FT — asked for a 90-day review period."}],[{"start":84.15,"text":"David Sacks, the Silicon Valley venture-capitalist and Trump adviser who had been pushing for light-touch scrutiny, said on X that the executive order was a “game changer” because it would allow AI labs “to comply with the voluntary framework without delaying new model releases”. "}],[{"start":100.30000000000001,"text":"He added he believed the order would “apply only to models that represent a meaningful step-change in cyber capabilities (eg Mythos), not to incremental version numbers of existing models”."}],[{"start":111.05000000000001,"text":"Tuesday’s order also asks AI companies to “collaborate with the federal government to select trusted partners that will have early access to covered frontier models to promote secure innovation and strengthen the cyber security of critical infrastructure”."}],[{"start":125.55000000000001,"text":"US officials have previously said the government would lean on the cyber expertise of large banks and financial institutions to vet frontier models, and perhaps also enlist the help of experts in other industries."}],[{"start":136.65,"text":"The order’s signing comes amid infighting over AI regulation within Trump’s orbit. Some of the president’s allies, such as Steve Bannon, have called for leading models to be brought under US government control while Sacks and others have warned against measures that might limit the technology’s growth."}],[{"start":154.15,"text":"At one point, Kevin Hassett, director of the National Economic Council, suggested frontier models ought to be subject to the same kind of approval process as pharmaceuticals and released only “after they’ve been proven safe, just like [a Food and Drug Administration] drug”."}],[{"start":169.75,"text":"His comments were met with fierce opposition from AI founders and investors, including some close to the administration, who argued such a regime would hamper US innovation."}],[{"start":180.4,"text":"The Trump administration has subsequently stressed that the order is not intended to impose a new regulatory burden on AI companies."}],[{"start":188.55,"text":"“I want to emphasise this is not regulation,” Bessent said in advance of the original signing date last month. “We’re going to continue to dominate the AI race, and we’re going to do it with the security at the forefront, which we think this executive order accomplishes.”"}],[{"start":203.9,"text":"However, Dean Ball, a former Trump adviser who last year wrote the administration’s AI Action Plan, called the order a “fairly major win for the safety contingent within the [administration]” and suggested it was “clearly teeing up the infrastructure for a model licensing regime”."}],[{"start":227.54999999999998,"text":""}]],"url":"https://audio.ftcn.net.cn/album/a_1780448834_9010.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。

马克•扎克伯格是不是太贪心?让股东来决定

如果这位亿万富翁创始人在意人们的看法,就应该说服其他投资者支持他的愿景。

AI监管不能照搬银行业模式

金融监管体系历经一个世纪的拼凑修补才逐渐成形。AI需要一套针对自身特点设计的监管机制。

FT社评:星链风波凸显欧洲亟须在太空领域迎头赶上

美国科技遥遥领先,但布鲁塞尔仍需建立自己的卫星网络并具备发射能力。

新加坡寻求追究相关公司责任,应对十年来最严重空气污染

政府威胁将对10家企业采取法律行动,称这些企业是邻国印尼山火的幕后推手,而山火造成的烟霾正在困扰新加坡。

美债收益率冲击下,欧洲必须自寻出路

埃里安:欧洲不能只盼着大西洋彼岸的借贷成本回落,它必须着手解决自身问题。

AI与通胀:央行行长该担忧到什么程度?

凯恩斯:这取决于一个价值万亿美元问题的答案:AI是否真能带来预期的生产力提升?
设置字号×
最小
较小
默认
较大
最大
分享×