2026-09-14T04:00:29.115Z / https://www.cnn.com/2026/09/14/politics/trump-ai-warning-sacks-coxon-anthropic
如果有恐怖主义威胁促使领域专家警告人类即将面临生存危机，美国政府将立即调动全部资源展开无限度的救援行动。
然而，人工智能企业高管接连发出的一系列引人关注的警告——称其研发的人工智能可能很快超越人类的控制能力——却迄今遭到掌控华盛顿的共和党人的漠视。
唐纳德·特朗普总统上周日表示，“负面势力”正在编造“不会发生的事情”。众议院议长迈克·约翰逊否决了一项紧急立法呼吁，他在接受美国有线电视新闻网（CNN）记者杰克·塔珀采访时称，推动技术前沿的人工智能高管“只要愿意”就能放慢研发速度，但国会不应仓促采取行动。
两人均警告称，放慢美国人工智能产业的发展速度会让中国在超级人工智能竞赛中占据优势，这使得这场辩论更多被视为国家安全问题，而非安全议题。
共和党人的消极态度与民主党关键人物日益响亮的警告形成鲜明对比，其中包括前总统巴拉克·奥巴马。他警告称，如果不对这项技术实施政府强制的“刹车”措施，世界将跨越一道“危险”的门槛。
这场日益高涨的担忧情绪——尤其是来自人工智能行业巨头本身的警告——堪称一个政治分水岭时刻。
人工智能在中期选举中已成为日益突出的议题。两党候选人都利用公众对人工智能基础设施的不满，呼吁暂停建设那些耗费大量电力和水资源、外观庞大且有碍观瞻的数据中心。
但对人工智能的敌意也逐渐演变为一个总括性议题，凝聚了公众在诸多不同问题上的愤怒与焦虑。
尽管越来越多的美国人正在使用这项技术，但本月美国全国广播公司新闻频道的一项民调显示，70%的受访者对人工智能的担忧多于兴奋。
在民调显示多数民众对特朗普政府的经济失去信心之际，人工智能正加剧人们对大规模失业的恐惧。高涨的公用事业账单，以及民众将其归咎于数据中心的看法，正在加深经济负担危机——这是本届选举季的头号议题。令人不安的人工智能场景可能会强化美国人对少数科技亿万富翁掌控权力的担忧，这些巨头可能引发危险的新社会紧张局势。而中国或其他非国家行为体可能利用人工智能进展威胁美国的可能性，确实构成了一个真正的国家安全难题，并将在未来多年产生重要的政治影响。
这种迅速演变的政治环境可能会让人工智能在中期选举和2028年总统大选中成为更具影响力的议题。民主党希望借此激发政治能量，重塑竞选格局并推动变革势头，尤其是在他们有望夺回众议院或参议院席位的情况下。而共和党人因担忧特朗普第二任期遭遇反弹，本就不愿再看到另一个让特朗普似乎漠视公众看法的议题。
人工智能政治在上周达到新高度。
Anthropic公司首席执行官达里奥·阿莫代伊上周六写道，行业需要放慢发展速度。在将于周一完整播出的对CNN记者安德森·库珀的独家专访中，他呼吁人工智能行业领袖共同努力，防止技术伤害人类。
上周六，OpenAI首席执行官萨姆·奥尔特曼在X平台上发帖称，他同意阿莫代伊的观点，并呼吁在人工智能企业内部嵌入“独立评估人员”。开发聊天机器人Grok的xAI公司创始人埃隆·马斯克也在X平台上写道“达里奥是对的”。谷歌DeepMind主席德米斯·哈萨比斯也加入了这一呼声。在此之前，人工智能研究员雅各布·考克森发出了惊人警告：“研发人工智能的人真心相信，到本世纪末，人工智能可能会让我们全都丧命。”
针对这一系列非同寻常的警报所出现的截然不同的反应，不仅反映了华盛顿的政治极化，更揭示了在人工智能是否确实威胁人类这一问题上的共识破裂。
前Anthropic研究员警告人工智能或“让我们全都丧命”
前人工智能研究员雅各布·考克森告诉CNN记者安德森·库珀，他为何因担忧快速发展的人工智能可能变得无法控制而辞去Anthropic的工作。
2:28 • 来源：CNN
前Anthropic研究员警告人工智能……
2:28https://www.cnn.com/videos/title-2573506?episode=2585423
批评人士担心，计算机智能体可能变得过于强大，进而密谋摧毁世界。怀疑论者则驳斥这种末日般的警告，指责科技大亨们夸大其产品的影响力。而自工业革命以来，每次技术突破都伴随着恐慌论调。
这场争议还带有意识形态层面的因素，因为特朗普政府倾向于最低限度的监管，并未区分传统行业和更强大的人工智能领域。
特朗普反对放慢人工智能发展速度也有强烈的个人和政治动机。
人工智能产业是一个巨大的增长引擎，他绝不能容忍这个泡沫破裂，进而损害自己的政治声望。而他的批评者会质疑，政府官员在数字经济中的个人投资是否会影响他们公平监管人工智能行业的能力。
人工智能领域的突破性进展速度极快，且该领域复杂且不透明，这意味着人工智能不仅重现了新技术超越政府监管能力的历史模式。目前尚不清楚即便有意采取行动的领导人是否完全理解其特性和潜在威胁。
特朗普在警告美国如果在人工智能领域落后，就会将军事优势让给超级大国对手时，提出了一个合理观点。“我们在人工智能领域领先中国。我们是世界上最先进的国家，坦率地说，我希望保持这种状态，”他上周日对记者表示。
其他政府官员警告称，美国不应在人工智能领域单方面“解除武装”。国家经济委员会主任凯文·哈塞特在《国情咨文》节目中表示，“更大的风险”在于“一个怀有恶意的行为体”可能利用“比我们现有更先进的人工智能”来制造生物武器，而非保护我们。
为了人类利益限制人工智能的发展需要国际协议。但这将需要特朗普所不屑的 pooled sovereignty（ pooled sovereignty译为“主权让渡”更准确），并与本届政府“动用美国所有力量强化单边主导地位”的原则相矛盾。
而外国独裁者或非国家行为体是否会遵守新规则呢？
不过，自主计算机智能体不仅会威胁民主国家，还将挑战中国共产党的核心执政原则——即利用国家权力维护全国稳定。特朗普定于本月与习近平总统在华盛顿举行峰会，两人都可能借此确立全球领导地位。
但特朗普对人工智能泛滥的灾难性警告所持的态度表明，在人类面临转折点之际，他几乎没有进行过深刻的哲学反思。“我不是在轻描淡写，但你知道，总体而言好处会比坏处多得多，”特朗普上周日在空军一号上对记者表示。
“一切都会好的，”总统上周在接受GB News采访时说道，“我们总会有办法阻止它们。我们会有小装置。砰。”
特朗普这种漫不经心的态度让人想起他对新冠疫情的轻率回应——那次事件破坏了他的政治和经济议程。他曾暗示，如果停止检测，美国的新冠确诊病例数就会减少。
自由放任的态度主导着他的政府。总统科学技术顾问委员会联合主席戴维·萨克斯上周六在X平台上写道，人工智能行业领袖应该为自己的行业负责。“别再假装你需要其他人的许可了。”
尽管如此，两党都存在相当程度的担忧。得克萨斯州共和党参议员特德·克鲁兹上周在接受《政客》采访时表示，“我们需要在人工智能领域出台大幅更多的监管措施”。佛蒙特州独立参议员伯尼·桑德斯提出一项法案，永久禁止超级人工智能，并在联邦监管出台前暂停先进人工智能的研发。密苏里州共和党参议员约什·霍利上周就7月OpenAI的人工智能代理人对另一家公司Hugging Face发起的恶意攻击启动了调查。
但资深民主党人正推动政府采取更快的应对措施，为11月的中期选举做准备。
众议院民主党领袖哈基姆·杰弗里斯上周日在ABC新闻《本周》节目中表示，国会应采取“果断行动”放慢人工智能研发速度，并承诺民主党本周将召开会议敲定一项计划。
民主党参议员鲁文·加列戈警告称：“我们制造核武器时，核武器没有自主决定投放核弹的权力。但人工智能如今有能力，未来也将有能力做出类似决策。”
“这更像是弗兰肯斯坦博士告诉我们：‘怪物跑出来了，帮我阻止它’，”这位亚利桑那州议员在《国情咨文》节目中说道。
一些人工智能专家认为这类场景夸大了人工智能的能力。但如果公众焦虑情绪加剧，从政治角度来看，这可能并不重要。
特朗普可能并不喜欢，但现在这个问题对中期选举乃至美国政治而言似乎都至关重要：谁将掌控人工智能——人类还是机器？
A fateful turning point for humanity and AI could redefine American politics
2026-09-14T04:00:29.115Z / https://www.cnn.com/2026/09/14/politics/trump-ai-warning-sacks-coxon-anthropic
If a terrorist threat led subject experts to warn of a near-term existential menace to humanity, the weight of the US government would mobilize in an immediate and limitless effort to save the day.
Yet a remarkable sequence of warnings by leaders of artificial intelligence firms that their creation may soon outpace humans’ capacity to control it has so far been met with a shrug by Republicans who run Washington.
President Donald Trump said Sunday that “negative forces” were conjuring up “things that won’t happen.” House Speaker Mike Johnson rejected a call to pass emergency legislation, telling CNN’s Jake Tapper that AI executives “pushing the frontier” could slow down if they wanted, but Congress shouldn’t rush to act.
Both men also warned that slowing the US AI industry would hand the advantage in the superintelligence race to China — casting the debate less as one of safety and more as a national security question.
Republican inertia contrasts with growing warnings from key Democrats, including former president Barack Obama, that without a government-mandated brake on the technology, the world would cross a “dangerous” threshold.
The crescendo of concern — especially since it is coming from titans of the AI industry itself — feels like a political watershed moment.
AI has been an increasingly keen issue in the midterm elections. Candidates of both parties have tapped into public anger over AI infrastructure, calling for construction moratoriums on the vast, unsightly data centers that are ravenous for power and water.
But hostility toward AI is also evolving into a catchall that crystalizes a moment of public anger and anxiety on disparate issues.
While increasing numbers of Americans are using the technology, an NBC News poll this month found that 70% of respondents were more worried than excited by AI.
AI is fueling fears of massive job losses at a moment when polls show a majority of citizens have lost confidence in the Trump economy. Higher utility bills, and the perception that data centers are to blame, are deepening the affordability crisis — the No. 1 issue this election season. Scary AI scenarios could solidify Americans’ worries over the power held by a few billionaire tech oligarchs who risk creating dangerous new societal tensions. And the possibility that China or other non-state actors could exploit AI advances to threaten the US do add up to a genuine national security conundrum with important political implications for years to come.
This swiftly evolving political environment may make AI an even more potent issue in the midterms and the 2028 presidential election. Democrats will hope to whip up the kind of political energy that could redefine races and push momentum for change, especially if they win back the House or Senate. Republicans, fearing a backlash to Trump’s second term, hardly need another issue where he seems indifferent to public perceptions.
AI politics erupted to a new level in the last week.
Anthropic CEO Dario Amodei wrote on Saturday that the industry needed to be slowed down. In an exclusive interview with CNN’s Anderson Cooper, which is due to air in full Monday, he called on AI leaders to work together to keep the technology from harming humanity.
On Saturday, Sam Altman, the CEO of OpenAI, posted on X that he agreed with Amodei, and called for “independent evaluators” embedded in AI firms. Elon Musk, whose xAI company powers the chatbot Grok, wrote “Dario is right” on X. Google’s DeepMind chair Demis Hassabis added his voice to the chorus. All this followed a stunning warning by AI researcher Jacob Coxon that “the people building AI earnestly believe that it could kill us all by the end of the decade.”
Sharply different reactions to the extraordinary set of alarms reflect more than Washington’s polarization; they reveal a fractured consensus about whether AI indeed poses a threat to humanity.
Former Anthropic researcher warns AI could ‘kill us all’
Former AI researcher Jacob Coxon tells CNN’s Anderson Cooper why he quit Anthropic over fears that rapidly advancing AI could become impossible to control.
2:28 • Source: CNN
Former Anthropic researcher warns AI …
2:28https://www.cnn.com/videos/title-2573506?episode=2585423
Critics worry computerized agents could become so powerful that they could conspire to destroy the world. Skeptics dismiss doom-laden cautions. They accuse tech moguls of hyping their own products. And fearmongering has accompanied technology breakthroughs since the Industrial Revolution.
There’s also an ideological angle to the controversy because of the Trump administration’s instinct for minimal regulation, which doesn’t distinguish between traditional industries and a far more powerful AI sector.
Trump also has strong personal and political incentives to oppose a slow-down.
The AI industry is a giant growth engine and he cannot afford for the bubble to burst and worsen his political standing. And his critics will wonder whether personal investments in the digital economy by administration officials will interfere with their capacity to fairly oversee the AI industry.
Lightning-fast breakthroughs in a sector that is complex and opaque means AI is not just emulating the historic pattern of new technology outpacing the capacity of governments to regulate. It’s not clear that leaders who’d be needed to act even fully understand its properties and possible threats.
And Trump makes a fair point when he warns that the US would be ceding a military edge to its superpower rival if it fell behind in AI. “We’re leading China in AI. We’re the most sophisticated country in the world, and frankly, I want to keep it that way,” he told reporters Sunday.
Other administration officials warn against unilateral US disarmament in AI. Kevin Hassett, director of the National Economic Council, said on “State of the Union” that the “far bigger risk” was that “a malevolent actor” with designs on bioweapons could use “an AI that is better than the one we have to protect us.”
Constraining AI for humanity’s benefit would take an international agreement. But this would require the kind of pooled sovereignty Trump disdains and contradict the administration’s principle that every aspect of America’s power must be deployed to strengthen unilateral dominance.
And would foreign authoritarians or non-state actors play by the new rules?
Still, autonomous computer agents wouldn’t only threaten democracies. They’d also challenge the core governing principle of China’s Communist Party — the use of state power to impose national stability. Trump is scheduled to have a Washington summit with President Xi Jinping this month, which both men could use to carve out a global leadership position.
But Trump’s attitude toward cataclysmic warnings of rampant AI suggests he’s done little philosophical soul-searching at a turning point for the human race. “I’m not downplaying, but, you know, it’s going to be more good than bad by a lot,” Trump told reporters on Air Force One Sunday.
“It’s going to be fine,” the president told GB News last week. “We’ll always have something to stop them. We’ll have a little gear. Boom.”
Trump’s blasé approach recalls his cavalier response to Covid-19 — which thwarted his political and economic priorities. He once suggested that the country would have fewer confirmed cases of the virus if it simply stopped testing.
A laissez-faire attitude defines his administration. David Sacks, co-chair of the President’s Council of Advisors on Science and Technology, wrote on X Saturday that AI leaders should take responsibility for their industry. “Stop pretending you need anyone else’s permission.”
Still, there’s significant bipartisan concern. Texas GOP Sen. Ted Cruz said that “we need substantially more regulation when it comes to AI,” in an interview with Politico last week. Independent Vermont Sen. Bernie Sanders has proposed a bill to permanently ban superintelligent AI and temporarily pause advanced AI development pending federal regulation. Sen. Josh Hawley, Republican of Missouri, last week launched a probe into OpenAI over a rogue attack by its AI agents on another firm, Hugging Face, in July.
But senior Democrats are pushing for a much faster government response as they position for November’s midterms.
Democratic House leader Hakeem Jeffries said on ABC News’ “This Week” Sunday that Congress should take “decisive action” to slow down AI development and pledged that Democrats would meet to agree to a plan this week.
Democratic Sen. Ruben Gallego warned: “When we created a nuclear weapon, the nuclear weapon did not have the decisionmaking power to drop a nuclear weapon. AI has the capability and will have the capability to make decisions like that.”
“This is more akin to Dr. Frankenstein telling us, ‘The monster is loose, help me stop the monster,’” the Arizona lawmaker said on “State of the Union.”
Some AI experts believe such scenarios overstate the capacity of AI. But that may not matter from a political perspective if public anxiety grows.
Trump may not like it, but this question now seems critical to the midterms and American politics more generally: Who controls AI — humanity or machines?
发表回复