Author: Nita Farahany Source: https://nitafarahany.substack.com/p/us-ai-policy-myths-and-realities Published: November 17, 2025
This is the US AI policy installment (Class 22 of 27) of Nita Farahany's AI Law and Policy course, published November 17, 2025. The class is anchored on the Trump AI Action Plan of July 2025 ("Winning the Race") and on Professor Alondra Nelson's argument that the administration's apparent deregulation is in fact selective intervention — what Nelson describes as the most government leaning-into industry "since the early 20th century." The special edition draws on Nelson's video and an American Enterprise Institute (AEI) roundtable analysis. It is organized around four claimed myths about US AI policy and a case study of the Anthropic–Sacks conflict.
The AI Action Plan's three pillars
The class summarizes the AI Action Plan as resting on three pillars.
The first, Accelerate AI Innovation, covers support for open-source and open-weight models; a National AI Research Resource buildout; AI Centers of Excellence; Section 132 tax-free training; Department of Labor retraining programs; regulatory reduction (a Request for Information on AI-hindering regulations); revisions to the NIST AI RMF to "eliminate references to misinformation, DEI, and climate change"; and CCP-alignment research on Chinese frontier models.
The second, Build American AI Infrastructure, covers new NEPA Categorical Exclusions for data centers; FAST-41 expedited permitting; nationwide Clean Water Act Section 404 permit consideration; opening federal lands for data centers; a CHIPS Program Office focus on return on investment; the Three Mile Island restart; small modular nuclear; and an AI-ISAC for threat sharing.
The third, Lead in International AI Diplomacy and Security, covers full-stack American AI export packages; chip location-verification features; international standards advocacy; and biosecurity measures, including nucleic acid synthesis screening for federally-funded research.
The four myths
Myth 1: US AI policy is all about deregulation
Per Nelson, "you have fallen for the mirage of deregulation." The class catalogs selective intervention across multiple fronts: trade warfare around critical minerals; aggressive semiconductor chip export controls; $100,000 H-1B visa fee proposals; and equity buybacks in companies such as Intel, where the government took a 10% stake. Nelson's characterization is that this is not laissez-faire but industrial policy.
Myth 2: The plan is clear and comprehensive
Drawing on the AEI roundtable, the class cites several participants questioning the plan's implementability. Will Reinhardt described building a spreadsheet to track the plan's 90+ action items because it specifies no lead agencies and no timelines. Miles Brundage offered "big if true," and argued that strict political vetting filters out talented researchers. Molly Kinder noted that "there is no Department of Education" to implement the workforce provisions. The class describes the AI Exports Program rollout in October 2025 as "extremely underwhelming" and "chaotic," with Commerce releasing a Request for Information instead of clear guidance — asking industry what they want to know about a program that was supposed to already exist.
Myth 3: The administration speaks with one voice
The class points to Vice President Vance's February 2025 Paris speech ("stop with this hand-wringing about safety") and to the renaming of the US AI Safety Institute to the "AI Security and Innovation Institute," while noting that Vance's later language evolved to acknowledge worker concerns. It describes a winner-picking problem, in which relationships between specific firms and the administration create "ad hoc-ness... not available to everyone equally."
Myth 4: Industry unanimously supports deregulation
Per Nelson's review of RFI responses, the American Hospital Association and similar organizations want regulation and clarity, because they need rule harmonization for healthcare adoption. The class also argues that state legislatures are passing AI laws not out of a preference for regulation but to fill a federal vacuum.
The Anthropic–Sacks conflict case study
The class presents the conflict between Anthropic and White House AI czar David Sacks as an illustration of the four myths. Anthropic uniquely supported California SB 53. In a "Curve" speech, Jack Clark of Anthropic said, "we are the child from that story... when we turn the light on we find ourselves gazing upon true creatures." Sacks accused Anthropic of "running a sophisticated regulatory capture strategy based on fearmongering" and of "fueling a state regulatory frenzy." Dario Amodei responded that "Anthropic, the administration, and leaders across the political spectrum want the same thing." The class presents the episode as exposing the limits of the administration's "free speech" rhetoric, the winner-picking problem, the implementation gap, and the absence of industry unanimity.
Relationships
- part-of: Nita Farahany intro course series (Class 22 of 27)
- related: America's AI Action Plan, Alondra Nelson (planned), David Sacks, Anthropic, America's AI Action Plan
- previous: Inside My AI Law & Policy Class 21: Understanding How the EU Regulates AI (Farahany, November 2025) next: Inside My AI Law & Policy Class 23: When Silicon Valley's Effective Altruists Meet Washington's Export Controls (Farahany / Hamilton, November 2025)