On September 14, 2026, former U.S. Vice President Kamala Harris publicly called for a "responsible slowdown" in frontier AI development and urged Congress to pass legislation establishing a dedicated federal regulatory agency. She also pressed the Trump administration to advance international agreements with China and other countries to limit dangerous uses of AI and establish global testing standards. This is Harris's most explicit public statement on AI policy since leaving the administration.
Harris's core argument is that the current pace of frontier AI development has outstripped the regulatory system's ability to respond, and that governments need time to build guardrails to prevent the technology from escaping human control. She put forward three specific demands: congressional legislation, an independent federal regulatory agency, and a multilateral international treaty—each building on the last, from hard domestic rules to soft international constraints, forming a comprehensive vision of a governance architecture.
From Soft Commitments to Hard Legislation: The Escalation of Policy Demands Over Three Years
To understand the significance of Harris's statement, it must be placed within a three-year timeline. In November 2023, as vice president, she attended the AI Safety Summit at Bletchley Park in the United Kingdom. The Biden-Harris administration established the U.S. AI Safety Institute (US AISI) within the Department of Commerce to implement the NIST AI Risk Management Framework, tasked with developing guidelines, tools, benchmarks, and best practices for assessing and mitigating dangerous AI capabilities. At the same time, it pushed 30 countries to sign a U.S.-led political declaration on the responsible use of military AI, covering transparency norms, human rights obligations, and accountability mechanisms.
The goal of that 2023 declaration was to commit countries to "strong and transparent norms applicable to all stages of the military AI lifecycle." The signatories were overwhelmingly Western-aligned countries; strategic competitors such as China and Russia never participated. In her London speech, Harris warned of "AI-driven cyberattacks on an unprecedented scale and the threat of bioweapons designed by AI," but the policy tools of that time—the NIST framework, voluntary corporate commitments, and summit political declarations—were all soft constraints with no enforcement mechanism.
Three years later, Harris has upgraded from "soft multilateral frameworks" to demands for "hard federal legislation + international treaties," essentially an implicit rejection of the effectiveness of the previous phase's governance path: when voluntary commitments cannot restrain labs' acceleration and political declarations have no substantive impact on the U.S.-China competition landscape, the next step can only be mandatory legislation.
The Labs' Paradox: Calling for a Slowdown with One Hand, Pushing New Models with the Other
Harris's call is not an isolated event. In the same window, Anthropic CEO Dario Amodei and OpenAI CEO Sam Altman also successively called for "slowing the frontier," arguing that the current pace of development has created substantive safety risks and supporting the creation of an independent safety evaluation mechanism.
These labs calling for a slowdown continued to release next-generation flagship models intensively during the same period. The posture of "calling for a slowdown with one hand while accelerating deployment with the other" reflects the logic of regulatory competition: whoever first pushes a regulatory framework into place has the opportunity, by setting compliance thresholds and controlling evaluation standards, to convert its existing advantages in technology and safety discourse into long-term barriers. For leading labs, "regulation" is to some extent also a competitive card.
For enterprise developers and AI users, the establishment of a federal regulatory agency means that future AI system deployments may need to undergo independent testing and mandatory risk assessment, similar to the pre-market approval process in pharmaceuticals. This will significantly increase compliance costs and lengthen the cycle from research and development to deployment. Small startups will bear disproportionate pressure, while large tech companies with full compliance teams may gain a relative advantage under the regulatory burden.
Trump's Rebuttal and Clearer Partisan Lines
The Trump administration's response to calls for a "slowdown" is outright rejection. Trump frames AI competition as the core track of great-power rivalry, emphasizes that "winning AI is winning everything," and dismisses the "slowdown" narrative as a political conspiracy to hold America back. This stance makes Harris's call to "advance international agreements with China and other countries" almost impossible to initiate operationally—under Trump's framework, any AI speed-limit agreement signed with China would be interpreted as a unilateral U.S. concession.
Former President Obama also expressed support for stronger AI regulation in the same window. He explicitly asked House Minority Leader Hakeem Jeffries to make AI legislation a top agenda item after retaking the House, and called on Democrats to propose a "very public dialogue framework" systematically covering issues such as AI safety, employment disruption, and child protection. The successive statements from Harris and Obama mean Democrats are consciously shaping AI regulation into a partisan brand for the 2026 midterm elections.
What Is Most Likely to Happen Next
The most likely short-term trajectory is that AI regulation will enter a partisan track at the congressional level rather than become bipartisan consensus legislation. Democrats will take Harris's and Obama's statements as their outline and build an election narrative around "federal regulatory agency + international treaty"; Republicans will counter with the logic that "AI competition comes first; regulation hinders innovation."
If Democrats retake control of the House in the midterms, the window for advancing federal AI regulatory legislation will open substantially; if Republicans hold their majority, the "full speed ahead" approach will continue through the 2028 election cycle. There will be almost no substantive progress on the international treaty front during Trump's term—the political preconditions for U.S.-China AI negotiations do not currently exist.
Whether an AI regulatory framework bill is formally introduced in the Senate and House within the next 90 days, and whether AI labs shift from "voluntary commitments" to "actively lobbying for legislation," will be key signals for judging the direction of the issue. If the latter occurs, it will indicate that leading labs have concluded that regulation is inevitable and are transforming from resisters into participants in rule-making.
© 2026 Winzheng.com 赢政天下 | 转载请注明来源并附原文链接