中期选举后企业面临更严峻的AI监管环境

民主党掌控国会意味着,企业应当预期各州对 AI 的监管方式将出现差异,基础设施建设也可能放缓。
Esther Shittu,AI Business 新闻撰稿人
Sanya Kushak via Getty Images
周二中期选举后政治风向的剧烈转变表明,AI 供应商和企业已在 Trump 总统任内习以为常的那种对 AI 不加干预的监管方式,很快将被打乱,或至少受到挑战。
选举清楚显示,美国民众表达了对数据中心的反感以及对 AI 的担忧,在许多州和县投票限制或叫停数据中心建设。
此外,在新近由民主党主导的国会中,一些议员承诺将传唤 AI 领袖宣誓作证,一群民主党民粹派人士则呼吁暂停 AI。新当选的密歇根州联邦参议员 Abdul El Sayed 呼吁对 AI 发展实施“全面暂停”。来自亚利桑那州的联邦众议员 Yassamin Ansari 主张反对 AI 超级智能。还有关于调查 AI 芯片出口管制的讨论,在 Trump 今年早些时候放松管制之后,可能会进一步收紧。
中期选举的结果反映了美国消费者对 AI 的情绪。这给股市带来了混杂信号:11 月 3 日选举次日,科技股和 AI 相关股票立刻大涨,而周四则大幅下挫。AI 芯片制造商 AMD、Micron 和 Nvidia 的股价下跌,助推了周四的跌势。
消费者不满,审查趋严
“选民不只是在为数据中心生气,他们还认为这是阻止 AI 的途径,”Constellation Research 创始人兼分析师 R “Ray” Wang 表示。“他们觉得 AI 会夺走他们的工作,而如果他们真能做点什么来阻止它,他们就想去这么做。”
Wang 认为,关于 AI 如何影响环境的错误信息,也促成了这一决定性的选举结果——选民以压倒性多数选择民主党,这是对 Trump 及其诸多政策前所未有的斥责。
然而,全国范围的投票结果应当提醒企业和供应商,将会出现“对 AI 公司更严格的审视,尤其是在安全、就业和社会方面,”RPA2AI Research 创始人兼分析师 Kashyap Kompella 表示。
尽管民主党很可能会在听证会、立法提案和调查中更关注这些问题,但对 AI 的联邦监管仍然遥遥无期。
“AI 监管正越来越像一片蛮荒西部,联邦、州和地方政府各行其是,”Kompella 表示,并补充说,在一个州可以接受的 AI 应用或工具,在另一个州可能就需要额外的披露、审计或监督要求。
企业应如何应对
企业应当预期各州将继续推动监管,伊利诺伊大学芝加哥分校负责数据与 AI 战略的副校监 Michael Bennett 表示。
“在监管方面,我们很可能会看到那里持续的行动,尤其是针对前沿模型,”Bennett 表示。“企业需要时刻保持警觉。”
一些可能影响 AI 模型的监管规定,或许会把重点放在确保这些模型不会对消费者及其隐私构成重大风险,也不会影响工人的就业——这是许多人担忧的问题。另一些监管规定则可能侧重于确保前沿模型的开发不会导致消费者面临更高的水电费和其他账单。
Kompella 表示,正确的应对方式是,企业应将 AI 治理视为核心业务实践,定期开展 AI 审计和风险评估,并建立问责机制。
此外,随着消费者对 AI 数据中心的强烈不满不断积聚,企业面临的局面更加复杂。美国民众,无论民主党还是共和党支持者,都投票支持对 AI 数据中心的建设施加更多限制。而在计划建设众多数据中心的得克萨斯州,现任共和党州长 Greg Abbott 已提议暂停部分现有的数据中心审批。
“我们至少会看到六个月的数据中心审批放缓期,”Wang 说,并补充称,这种放缓将损害美国在与中国的 AI 竞赛中的处境,并导致美国国内 AI 应用出现更大规模的收缩。
“过去数据中心建设可能不需要太多审查,而现在将面临如此繁多的监管繁文缛节,”他说。
“我们需要 AI 算力,同时也要在国家安全、地方管控和未来走向之间找到平衡,”Wang 继续说道。
由于对数据中心的敌意可能导致基础设施何时获批、以及企业能否获得合适算力的不确定性,Kompella 表示:“那些对 AI 做出重大投入的企业还需要考虑监管确定性、能源可获得性和基础设施韧性。”
Bennett 表示,对于那些与 Anthropic、OpenAI 和 Google 等大型前沿模型供应商有业务往来、依赖其服务并签有覆盖未来几年合同协议的公司而言,密切关注金融市场十分重要。
“随着这些调查几乎肯定会启动,很难想象哪家前沿公司会就此消失,但你完全可以想象,国会的调查行动会导致它们政策出现调整和变化,”他说。由于需要大量第三方或政府审查,这些变化可能拖慢模型发布。
市场对周二的中期选举作出了剧烈波动的反应。在 Trump 于 11 月 5 日(周三)表示将与国会合作后,道琼斯指数飙升 1,500 点。这是自 4 月以来最大的单日涨幅,标普上涨 1.4%。周四则情况不同,随着投资者重新评估民主党在选举中大胜所释放的信号,三大股指大幅下挫。
不过,也有人认为 AI 监管可能收紧存在积极一面。Kompella 表示,监管增多可能促使一些州通过提供有吸引力的监管环境、能源基础设施和激励措施来争夺 AI 产业投资。
“各州之间的竞争有助于确定哪些监管方式最为有效,”他说。Kompella 补充道,另一个影响是,旨在让大型 AI 公司承担责任的监管,也可能反而使它们变得更强大。
“规模更大的公司有资源消化合规成本、应对复杂的监管要求,并转移基础设施投资,”Kompella 说。“规模较小的公司和初创企业可能会觉得难度大得多。政策制定者需要在保护公众与维护竞争之间取得平衡。”
AI Business 新闻撰稿人
Esther Shittu 自 2021 年起报道 AI 技术与行业趋势。作为 Targeting AI 播客的联合主持人,她与专家、思想领袖和从业者对话,探讨 AI 领域的关键进展。加入 AI Business 之前,她曾为 SearchEnterpriseAI、New York Daily News、Bklyner 和 Brooklyn Daily Eagle 撰稿。不深入钻研 AI 世界的时候,她会把时间投入自己热爱的项目,并抚养三个女儿。
想在 Google 搜索结果中看到更多 AI Business 的报道吗?
译文由英文原文机器翻译,术语按行业惯例统一,可在文末展开原文核对。
查看英文原文(Enterprises face tougher AI regulatory landscape after midterms)
Enterprises face tougher AI regulatory landscape after midterms
Enterprises face tougher AI regulatory landscape after midterms
A Democratic-controlled Congress means companies should expect differences in how states regulate AI and possibly a slowdown in infrastructure buildout.
Esther Shittu,News Writer, AI Business
Sanya Kushak via Getty Images
A dramatic shift in the political tide following Tuesday's midterm elections indicates that the hands-off regulatory approach to AI that AI vendors and enterprises have become accustomed to under President Trump will soon be disrupted or at least challenged.
The elections clearly showed American citizens expressing their dislike of data centers and fears about AI, voting to restrict or stop data center development in many states and counties.
Moreover, some lawmakers in the newly Democrat-dominated Congress are promising to bring AI leaders to testify under oath, and a bloc of Democratic populists are calling for an AI pause. Abdul El Sayed, the newly elected U.S. senator from Michigan, has called for a “full-on pause” on AI development. A U.S. representative from Arizona, Yassamin Ansari, is advocating against AI superintelligence. There is also talk of investigations into AI chip export controls, potentially to tighten them after Trump relaxed controls earlier this year.
The results of the midterms reflect consumer sentiment about AI in the U.S. They led to mixed signals for the stock market, with an immediate surge in tech and AI-related stocks the day after the November 3 elections and a sharp decline on Thursday. Declines in shares of AI chipmakers AMD, Micron and Nvidia helped trigger Thursday’s downturn.
Upset consumers, greater scrutiny
“Voters were mad not just about data centers, but they thought it was the way of stopping AI,” said R “Ray” Wang, founder and analyst at Constellation Research. “They felt AI was going to take away their jobs, and if they could actually do something to stop it, they wanted to go do that.”
Wang argued that misinformation about how AI affects the environment also contributed to the decisive election results, in which voters resoundingly opted for Democrats, in an unprecedented rebuke of Trump and many of his policies.
However, the nationwide ballot results should alert enterprises and vendors that there will be “greater scrutiny of AI companies, particularly around safety, jobs and society,” said Kashyap Kompella, founder and analyst at RPA2AI Research.
While Democrats will likely give greater attention to these issues in hearings, legislative proposals and investigations, federal regulation of AI remains a far way off.
“It is increasingly a wild west of AI regulation, with federal, state and local governments moving in different directions,” Kompella said, adding that an AI application or tool that is acceptable in one state could require additional disclosure, auditing or oversight requirements in another.
How enterprises should respond
Enterprises should expect states to continue to drive regulation, said Michael Bennett, associate vice chancellor for data and AI strategy at the University of Illinois Chicago.
“We’re likely to see continued action there as far as regulation, especially for frontier models,” Bennett said. “Enterprises will need to keep their antenna up.”
Some regulations that could affect AI models might focus on ensuring they don’t pose significant risks to consumers and their privacy or affect workers’ jobs, a fear among many. Other regulations could focus on ensuring that the development of frontier models doesn’t lead to higher utility and other bills for consumers.
The right response is for enterprises to treat AI governance as a core business practice, conduct regular AI audits and risk assessments, and have accountability mechanisms in place, Kompella said.
Moreover, as intense consumer vitriol builds around AI data centers, enterprises face greater complexity. Americans, both Democrats and Republicans, voted for more restrictions on AI data center buildouts. And in Texas, home to many planned data centers, the current Republican governor, Greg Abbott, has proposed pausing some existing data center approvals.
“We’re going to see at least six months when data center approvals get slowed down,” Wang said, adding that a slowdown will hurt the U.S. in the AI race against China and cause a larger pullback of AI use in the country.
“What might have taken less scrutiny for a data center to be built is now going to have so much regulatory red tape,” he said.
“We need AI capacity, and you want to find the balance among national security, local control and what the future looks like,” Wang continued.
Since the animosity against data centers can lead to uncertainty about when infrastructure will be approved and whether enterprises will have the right computing capacity, “enterprises that make significant AI commitments will also need to consider regulatory certainty, energy availability and infrastructure resilience,” Kompella said.
For companies that do business with large frontier model providers, such as Anthropic, OpenAI and Google, rely on their services and have contractual agreements covering the next few years, it's important to pay close attention to the financial markets, Bennett said.
“As these investigations almost certainly get underway, it’s hard to imagine that any of the frontier companies are going anywhere, but you certainly could imagine modifications, changes to their policies as a result of investigatory efforts of Congress,” he said. These changes could perhaps slow model releases due to extensive third-party or government review.
The market has responded with dramatic volatility to Tuesday’s midterms. Following Trump’s remarks Wednesday, Nov. 5, that he would cooperate with Congress, the Dow Jones shot up 1,500 points. It was the largest single-day gain since April, and the S&P spiked by 1.4%. The story was different on Thursday, when the three major stock market indexes plunged as investors reassessed signals from the Democratic election sweep.
Some see an upside, though, to the likely tightening of AI regulation. More of it could lead some states to vie for AI industry investment by offering attractive regulatory environments, energy infrastructure and incentives, Kompella said.
“Competition among states can help establish which regulatory approaches work best," he said. Another effect, Kompella added, is that regulations meant to hold large AI companies accountable could also strengthen them.
“Bigger companies have the resources to absorb compliance costs, navigate complex regulations and relocate infrastructure investments,” Kompella said. “Smaller companies and startups may find it considerably harder. Policymakers will need to strike a balance between protecting the public and preserving competition.”
News Writer, AI Business
Esther Shittu has covered AI technologies and industry trends since 2021. As co-host of the Targeting AI podcast, she talks with experts, thought leaders and practitioners exploring critical AI developments. Before AI Business, she wrote for SearchEnterpriseAI, the New York Daily News, Bklyner and the Brooklyn Daily Eagle. When she's not diving deep into the world of AI, she spends her time on passion projects and raising her three daughters.
Want more AI Business stories in your Google search results?
原文为英文,正文由自动抓取并保留原文。
原文发布于 AI Business(Enterprises face tougher AI regulatory landscape after midterms),标题、摘要与正文由机器翻译,术语按行业惯例统一,未经人工逐句校订。
