CLS News, May 22 (Edited by Xiaoxiang) — U.S. President Donald Trump on Thursday (May 21) delayed signing an executive order that would have granted the government greater regulatory authority over the AI industry, saying he did not want to take any action that could slow America’s pace in the global AI race.
Hours before a planned signing ceremony with industry executives in attendance, Trump announced the delay in the White House Oval Office. He told reporters the order would have required AI companies to preview their models with the federal government — a step he argued would put the U.S. at a disadvantage against other nations in AI competition.
“I do think that could be a hindrance, and I want to make sure it’s not,” he said, referencing the jobs AI is creating. “I don’t want to do anything that impedes that lead.”
The move is the latest sign of division within the White House over AI policy.
As powerful models like Anthropic’s Mythos have emerged, administration officials have pushed to bolster cybersecurity — but some AI advisors have warned such measures could slow the development of new AI tools, creating two factions locked in debate over the tech agenda.
Notably, Trump’s decision came just hours after California’s Democratic Governor Gavin Newsom issued a state executive order requiring policies to address job losses and economic disruption caused by AI — issues fueling growing public pushback against the technology.
A Temporary Victory for the “Pro-AI” Camp
Trump’s decision marks a win for business-oriented advisors, including venture capitalist David Sacks. Since the start of Trump’s second term, Sacks has championed industry-friendly AI policies, arguing the economic benefits of AI models outweigh their risks.
A senior administration official recently indicated the White House was considering AI regulations modeled after the U.S. Food and Drug Administration (FDA)’s drug approval system.
Sacks, however, said on a recent podcast he co-hosted that Trump and senior officials oppose an FDA-style approval process for AI — a prospect that has also alarmed Silicon Valley, which fears it could delay model development.
“There’s a very visceral negative reaction because we know how damaging that would be to innovation,” he said.
Sacks currently serves as co-chair of the President’s Council of Advisors on Science and Technology (PCAST) and previously headed AI and cryptocurrency policy — widely known as the “AI czar.”
AI Security Risks Demand Urgent Action
Nevertheless, many industry insiders continue to warn that action on AI security cannot wait.
The push for stricter regulation gained steam after Anthropic’s Mythos model alarmed senior White House officials and business leaders, who feared its exceptional precision in identifying software vulnerabilities could enable cyberattacks. OpenAI has begun previewing similar models for clients and, like Anthropic, has restricted access at the government’s request.
The delayed executive order would have asked companies to voluntarily share advanced models with the government. It also would have required national security and cybersecurity officials to collaborate with agency heads and top tech firms to address software flaws uncovered by models like Mythos.
The goal was to protect small-town hospitals, banks and other entities that lack full access to cutting-edge technology or the resources to build robust cyber defenses — a top priority for officials including Vice President Vance and Treasury Secretary Bessent.
“We are striving to strike a balance between security and innovation,” Vance said at a press briefing Tuesday. “We believe the Trump administration has struck the right balance here, but it’s something we must continue to work on.”
A recent survey by the conservative-leaning Institute for Family Studies (which supports a stronger AI regulatory framework) found over 80% of registered U.S. voters back pre-release testing for AI models. Recent polls also show public support for slowing AI development and limiting data centers that power the technology — though the administration maintains AI’s economic benefits outweigh such concerns.
The breakneck pace of AI advancement has not only challenged U.S. officials but also increased pressure on Congress to pass legislation. Trump previously proposed a framework that, according to the administration, includes robust oversight for businesses and consumer protections — though many Democrats and Republicans argue the proposal is too weak to become law.
Source: CLS News
最后关头“放鸽子”！特朗普推迟签署AI安全行政令 背后有何考量？
财联社5月22日讯（编辑 潇湘）当地时间周四(5月21日)，美国总统特朗普推迟签署了一项本将赋予政府对AI行业更多监管权的行政命令，他表示不希望采取任何可能会拖慢美国在AI竞赛中步伐的行动。
在原定于周四举行的、计划邀请行业高管出席的签字仪式开始前几小时，特朗普在白宫椭圆形办公室宣布了这一行动的推迟。他对记者表示，该行政命令原本将要求人工智能公司向联邦政府预先展示其模型，此举将使美国在与世界其他国家的AI竞争中处于劣势。
“我确实认为这可能会成为阻碍，而我希望确保不会如此，”他在提到AI技术正在创造的就业岗位数量时说道，“我不希望做任何可能阻碍这一领先优势的事情。”
此举是白宫在AI政策上存在分歧的最新例证。
随着Anthropic公司的Mythos等强大模型的发布，政府官员一直在努力加强网络安全，但一些AI顾问担心这些措施可能会减缓新人工智能工具的开发，从而在政府内部形成了就技术议程争执不休的两个阵营。
值得一提的是，特朗普做出这一举动前数小时，民主党籍的加州州长加文·纽森刚刚发布了一项州行政令，要求该州制定政策以应对人工智能造成的失业和经济混乱——这些问题正引发公众对AI技术日益强烈的抵制。
“亲AI”阵营的暂时胜利
特朗普的决定对以商业为导向的顾问们来说是一场胜利，其中包括风险投资家大卫·萨克斯(David Sacks)。自特朗普第二任期开始以来，萨克斯一直推动采取对行业友好的AI政策。萨克斯等人认为，AI模型的经济效益大于风险。
一位政府高级官员最近曾暗示，白宫正在考虑制定与美国食品和药物管理局(FDA)批准药品所用系统相似的AI监管规则。
对此，萨克斯近日在其联合主持的一场播客中表示，特朗普及高级官员不希望对AI采用类似FDA的审批流程，而硅谷也对该设想感到担忧，因为这可能会延缓模型开发。
“这里存在一种非常本能的负面反应，因为我们知道这会对创新造成多大的损害，”他说。
萨克斯现任总统科技顾问委员会联席主席，此前曾担任人工智能与加密货币事务负责人，也就是外界俗称的“AI沙皇”。
AI安全防控刻不容缓
然而，不少业内人士目前依然警告，对AI安全的防控措施已刻不容缓。
在此之前，由于Anthropic旗下的Mythos模型令白宫高层官员和商界领袖感到担忧——他们担心这些模型在识别软件漏洞方面过于精准，甚至可能被用于网络攻击——本届政府正考虑加强监管。OpenAI正在向客户预览类似的模型，并且与Anthropic一样，在政府建议下限制了访问权限。
那份被推迟的行政命令原本要求企业自愿向政府展示先进模型。它还要求国家安全和网络安全官员与各机构负责人及顶级科技公司合作，解决包括Mythos在内的模型所发现的软件漏洞。
其目标是保护那些无法全面获取顶尖技术或缺乏资源来建立复杂网络防御的小镇医院和银行等实体。保护这些实体是政府官员的优先事项，包括副总统万斯和财政部长贝森特。
“我们正在努力在安全与创新之间取得平衡，”万斯周二在新闻发布会上表示。“我们认为特朗普政府在这方面做到了恰当的平衡，但这仍是我们必须持续努力的方向。”
倾向保守派的智库Institute for Family Studies(该机构支持建立更多AI监管框架)最近的一项调查显示，超过80%的美国登记选民支持在AI模型发布前进行测试。近期民调显示，公众支持放缓AI发展并限制支撑AI模型运行的数据中心的提案，但政府认为AI带来的经济效益大于这些担忧。
人工智能发展的迅猛势头不仅给美国政府官员带来了挑战，也加大了国会通过相关立法的压力。特朗普此前曾提出了一套框架，据政府称该框架包含对企业的充分监管和对消费者的保护，尽管许多民主党人和共和党人认为该提案力度不足，难以成为法律。
来源：财联社
返回首页