内容来源：https://www.theverge.com/column/845955/donald-trump-big-tech-2025

内容总结：

【《观察者》年终特稿：硅谷如何将特朗普变为“科技寡头盟友”】
2025年曾被预测为特朗普与民粹主义右翼“驯服科技巨头”之年，然而现实走向截然相反。随着人工智能浪潮席卷全球，科技巨头正以空前速度重塑美国政治规则，甚至将前总统特朗普转变为自身利益的维护者。

在过去一年中，硅谷通过向特朗普关联机构“捐款”、推动 MAGA 网络意见领袖影响白宫决策、吸引亿万富翁以“金雕像”式示好等方式渗透权力核心。但更值得关注的是其对人工智能立法体系的激进改造：科技公司一边游说国会禁止各州制定本地AI法规，却未提出替代性联邦立法；在受挫后转而推动总统签署行政令，惩罚试图执行地方法规的州。他们甚至试图通过接管国会图书馆来修改版权法，并以“与中国竞争”为由游说华盛顿削弱AI监管。

与此同时，AI的社会代价正在显现——就业岗位快速被替代，青少年心理健康受生成式AI影响，数据中心耗能激增，AI武器化风险上升。尽管两党选民对AI的焦虑与日俱增，科技巨头的游说焦点始终集中于解除约束，而非构建应对现实风险的防护网。

值得玩味的是，曾借助“反科技巨头”民粹情绪上台的特朗普，如今正协助AI亿万富翁们进一步扩张权力。当普通民众在电台热线中追问“数字孪生”立法进展时，华盛顿的回应却始终停留在“科罗拉多州反偏见法案”这类边缘修补上。

随着2025年步入尾声，硅谷与政治权力的融合已超越传统游说范畴，演变为对法律根基的挑战。而在假日季的短暂停歇后，这场关乎技术、权力与人类未来的博弈，将在新年迎来更剧烈的碰撞。

（本期《观察者》休刊两周，将于1月6日恢复更新）

中文翻译：

欢迎阅读2025年最后一期《监管者》专栏。若您尚未订阅The Verge，请点击此处注册，确保自己不在2026年的"顽皮名单"上。若您已是我们的订阅用户——哎呀，您真是太贴心了。

硅谷如何将特朗普变为"兄弟寡头"一员
2025年本应是特朗普与民粹主义右翼挫败科技巨头的年份，但现实并未按剧本上演。

上周，我在WNYC电台的《布莱恩·莱勒秀》节目中探讨了关于特朗普总统试图禁止各州自行制定人工智能法律的报道。我很少参与广播节目，却格外珍视这样的机会——原因很独特。有线电视新闻只给90秒阐述观点，播客节目虽能与同行畅聊一小时，却容易陷入小圈子的内部讨论。而广播节目中，普通听众会直接来电提问，真切诉说报道内容如何影响他们的生活。这让我开始思考：在华盛顿那封闭的政治泡泡之外，世界究竟在发生什么？

这次有位女士来电询问：国会是否已着手制定关于"数字分身"的法律？这种能模仿人类行为的生成式AI模型正被企业用于客户服务；更广泛地说，具有自主行动能力的AI正以极低成本取代人类员工的工作。我迅速检索记忆，却想不起任何直接规范数字分身的州或联邦法律（科罗拉多州的反歧视法最接近，但仅针对AI在招聘决策中的应用而非后续使用）。

过去一年，我大量记述科技行业在华盛顿的政治博弈：企业通过向特朗普"非营利组织""捐款"规避游说限制、"让美国再次伟大"阵营的网络意见领袖左右白宫决策、埃隆·马斯克被卷入特朗普阵营肥皂剧式的权力游戏、亿万富翁们靠镀金雕像博取特朗普欢心……但最让我反复思考的，仍是人工智能的政治博弈——特别是科技行业如何挑战维系美国政府的珍贵传统，急速将政治风向扭转为对自己有利的局面。诚然，科技公司向官员开出巨额支票保其仕途，组建AI超级政治行动委员会，准备无限制打击支持严格AI监管的候选人，但这不过是常规政治手段。

不寻常的是他们激进且迅速地试图重塑法律体系——更准确地说，是消除任何可能约束他们的法律边界。他们试图让国会禁止各州制定AI法律，却未提出替代性联邦法案；失败后又说服总统签署行政令，惩罚试图执行本州法律的地区。他们企图接管国会图书馆以修改版权执法与知识产权保护，甚至抛出联邦接管理论：或许联邦通信委员会对电信的监管权可延伸至AI领域？更说服了不少华盛顿政客：为在AI竞赛中对抗中国，必须废除这些法律。

他们极少主动提议解决AI造成的切实且日益增长的人文代价。多项民调显示两党选民都对AI感到不安，工作岗位正被AI快速取代，几乎每天都有生成式AI对用户（尤其是青少年）造成心理伤害的新案例。这还未提及数据中心的环境影响、敌对势力对AI的武器化应用（是的，中国是其中之一），以及那些着眼长远者所担忧的AI生存威胁。

今年二月入职时（当时科技巨头CEO们刚见证特朗普就职，马斯克开始大幅裁减联邦雇员数周），我为The Verge的政治报道定下基调：技术正在改变人类行为，而人类行为塑造政治。当时我预计特朗普将代表席卷而来的民粹主义不满浪潮——这股主要针对科技巨头的浪潮将他推回总统宝座，他也理应代表这些民众的利益。

但不到一年，形势已然逆转：特朗普的选民们正面对抽象无形且不受约束的AI力量，这种力量正以超乎想象的方式渗透他们的生活——而总统却乐见其成，甚至协助创造AI的亿万富翁们掌控全局。

本周The Verge精选：

• 《喂养机器》（作者：乔什·齐扎 & 海登·菲尔德）：OpenAI和Anthropic等前沿实验室为争夺通用人工智能领先地位需要海量数据，这耗费数十亿美元资金，而Mercor、Handshake等鲜为人知的公司正借AI热潮获利。
• 《Stack Overflow用户不信任AI，却仍在用它》（解码器专栏）：CEO普拉桑特·钱德拉塞卡与总编辑尼莱·帕特尔探讨ChatGPT如何成为Stack Overflow的"生存危机时刻"。
• 《千页文件揭示DOGE多少秘密》（作者：劳伦·费纳）：尽管布伦丹·卡尔即将赴国会作证，新披露文件仍未明确说明DOGE在联邦通信委员会的具体作为。
• 《税收抵免到期前的太阳能板安装狂潮》（作者：贾斯汀·卡尔玛）：太阳能行业正艰难转型以应对特朗普对清洁能源的打击。
• 《家长呼吁纽约州长签署里程碑式AI安全法案》（作者：海登·菲尔德）：他们称该法案是"最低限度护栏"，应成为行业标准。
• 《AI芯片机架重得离谱》（作者：艾丽莎·威尔）：老旧数据中心物理上无法支撑成排的GPU，这正是大规模新建AI数据中心的原因之一。

节日休刊启事
《监管者》专栏将休假两周欢度节日，恰逢其时地于1月6日回归。最后奉上《虎胆龙威》编剧史蒂文·德苏扎对时局的标准论断：

谨以韦氏词典年度词汇的精神祝福各位：混沌圣诞快乐，混沌新年快乐。

英文来源：

Hello and welcome to 2025’s last issue of Regulator. If you’re not a Verge subscriber, get off the 2026 naughty list by signing up here. And if you’re a Verge subscriber — well, dang, that’s really nice of you.
How Silicon Valley turned Trump into a fellow broligarch
2025 was supposed to be the year Trump and the populist right humbled Big Tech. It didn’t work out like that.
How Silicon Valley turned Trump into a fellow broligarch
2025 was supposed to be the year Trump and the populist right humbled Big Tech. It didn’t work out like that.
Last week, I appeared on The Brian Lehrer Show on WNYC to talk about my reporting on President Donald Trump’s attempt to ban states from making their own AI laws. I don’t often get to appear on radio, but I love doing it for one unique reason. On cable news, you have 90 seconds to make your point and that’s all you get. On podcasts, you fall into a rhythm in a room of peers for an hour and even though it can be fun, it runs the risk of getting too insider-y. But on radio, regular everyday listeners get to call in, ask questions, and tell you exactly how the thing you’re reporting on impacts their lives. It makes you start thinking about what happens outside the weird little Washington bubble where your reporting comes from.
In this case, a woman called in to ask whether Congress had started working on any laws addressing “digital twins,” a generative AI model that mimics human behavior and is used by corporations for customer-facing interactions, and broadly, agentic AI, which is filling in — rather cheaply — the work that was once done by human employees. I had to quickly rack my brain to see if I’d run into any state or federal laws, drafts, or whatever that directly addressed the use of digital twins, and I couldn’t. (Colorado’s anti-bias laws come the closest, but address AI’s usage in employment decisions — not what happens afterward.)
Over the past year, I’ve written a lot about the tech industry’s version of Washingtonian political drama: companies skirting lobbying restrictions by “donating” to Trump’s “nonprofits,” MAGA internet influencers steering the White House’s policy decisions, Elon Musk getting dragged into Trumpworld’s soap opera-esque power plays, billionaires winning Trump’s favor one gold statue at a time. But the story I keep finding myself coming back to, again and again, is the politics of artificial intelligence — specifically, the industry’s attempts to swiftly turn politics in their favor, in a way that challenges the precious norms that’s kept the US government together. True, tech companies have signed massive checks for elected officials, promising to keep them in office, and created their own AI super PACs, preparing to spend unlimited amounts of money on targeting candidates promising unfavorable AI regulations. But that’s a normal way to play the political game.
What’s unusual is their aggressive and swift attempt to reshape the law altogether — or, rather, eliminate any law that would place a boundary on them. They’ve tried to get Congress to ban states from writing their own AI laws, without suggesting any federal law to replace them; when those attempts failed, they convinced the president to sign an executive order that would punish the states trying to enforce their own laws. They’ve tried taking over the Library of Congress in order to change copyright enforcement and IP protection and floated several theories for a federal takeover: Mayhaps the Federal Communications Commission’s authority over telecoms could give the feds the power to regulate AI? And they’ve convinced enough people in Washington that they need those laws removed in order to compete against China in the AI race.
Very rarely do they suggest anything that proactively addresses the immediate, real, and growing human cost of artificial intelligence. Several polls show a bipartisan nervousness around AI, jobs are being lost to AI at a rapid pace, and every day, it seems like a new story comes out about how generative AI has psychologically harmed its users — particularly its youngest ones. That’s to say nothing about the environmental impact of data centers, the weaponization of AI by adversarial actors (yes, China is one of those), and to those looking even further ahead, the “doomer” position that AI poses an existential risk.
When I first came on board in February — one month after Big Tech CEOs watched Trump get sworn into office, and weeks after Elon Musk began decimating the federal workforce — I laid out my thesis for The Verge’s political coverage: Technology is transforming human behavior, and human behavior shapes politics. At the time, I’d anticipated that Trump would represent the wave of populist discontent, largely targeted against Big Tech, that had swept him back into office, and that he’d represent their interests.
But less than a year later, it seems like the tables have turned: Trump’s voters are confronting the abstract, faceless, and unrestrained force of artificial intelligence influencing their lives in dimensions they could have never imagined — and the president is all too happy to help its billionaire creators take over.
This week at The Verge:

• “Feeding the machine”, Josh Dzieza & Hayden Field: Frontier labs like OpenAI and Anthropic need vast amounts of data in the race to achieve AGI. This comes at a pretty penny — billions of dollars — and little-known companies like Mercor and Handshake are cleaning up in this AI hype cycle.
• “Stack Overflow users don’t trust AI. They’re using it anyway”, Decoder: CEO Prashanth Chandrasekar talks to The Verge editor-in-chief Nilay Patel on how ChatGPT became an “existential moment” for Stack Overflow.
• “What 1,000 pages of documents tell us about DOGE”, Lauren Feiner: As Brendan Carr heads to Capitol Hill, newly released documents still don’t say much about what DOGE did at the FCC.
• “The ‘mad rush’ to install solar panels before tax credits run out” Justine Calma: The solar industry is pivoting to survive Donald Trump’s attacks on clean energy.
• “Parents call for New York governor to sign landmark AI safety bill”, Hayden Field: They called it “minimalist guardrails” that should set a standard.
• “Racks of AI chips are too damn heavy”, Elissa Weille: Old data centers physically cannot support rows and rows of GPUs, which is one reason for the massive AI data center buildout.
  And now, even more Holiday Season Recess.
  Regulator will be off for the next two weeks for the holidays, and quite fittingly, will return on January 6th. In the meantime, here is a canon position on The Discourse from Die Hard writer Steven de Souza:
  In the spirit of Merriam-Webster’s word of the year: Merry slop-mas and a happy slop year.