AI Policy Wiki
Dashboard

AI Policy (umbrella)

high confidence · updated 2026-06-06

Top-level umbrella for the wiki's coverage of AI policy — the deliberate set of governmental and institutional decisions about whether, when, where, and how AI is developed, deployed, and constrained. Indexes the wiki's specific governance, regulation, and policy-instrument coverage.

AI policy is the deliberate set of governmental and institutional decisions about whether, when, where, and how AI is developed, deployed, constrained, taxed, subsidized, and integrated into existing law. It is distinct in scope from AI governance, which is the taxonomy of governance modes: AI policy is what specific actors do, while AI governance describes how their actions are structured. Coverage of specific instruments and jurisdictions is split across legislation/, government/, entities/ (regulators), analysis/, and many concepts/ pages, indexed below.

Organization by instrument type

InstrumentFolderExamples
Federal legislationlegislation/LIFT AI Act (Literacy in Future Technologies Artificial Intelligence Act), America's AI Action Plan (EO-track)
State legislationlegislation/California SB 53, Colorado AI Act (SB 24-205), California SB 243 — Companion Chatbots, Tennessee SB 1580 / HB 1470 — AI Mental-Health Professional Impersonation Prohibition, Idaho SB 1297 — Conversational AI Safety Act, Nebraska LB525 — Conversational AI Safety Act + Agricultural Data Privacy Act etc.
Executive orders / EOslegislation/Executive Order 14365, EO — Trump Federal Preemption of State AI Laws (Dec 11, 2025)
International frameworkslegislation/EU AI Act (Regulation 2024/1689), Paris AI Action Summit Declaration (2025)
Agency rulemakingentities/ + legislation/BIS rules, FTC AI orders, FCC AI orders
Formal standardsstandards/ISO/IEC 42001, NIST AI RMF, NIST SP 800-234
Procurement frameworksconcepts/, entities/GSA OneGov USai, IL5/6/7 cohort decisions, Procurement-Driven AI Governance
Voluntary frameworksconcepts/, sources/RSP, Preparedness Framework, FSF
Litigation as quasi-policylitigation/, analysis/AI Copyright Litigation — Analysis, Musk v. Altman (and OpenAI / Microsoft / Brockman), Anthropic v. United States (Pentagon ban challenge)

Organization by jurisdictional layer

Organization by substantive issue

Substantive issues are indexed through the concepts/ folder. The umbrella concept page AI Governance (umbrella) enumerates 11 governance modes.

Institutional-design debate: "fail gracefully"

In a May 12, 2026 essay in Asterisk Magazine (the "Dagger" column), the author argues that AI policy must "fail gracefully" — that it should be designed to survive an unaligned executive branch rather than depending on one. The essay cites the Trump administration's reported retaliation against Anthropic over autonomous-weapons and surveillance guardrails as evidence that executive-dependent AI governance is, in the author's terms, brittle, and proposes three structural design principles:

  1. Statutory whistleblower protections with a 180-day federal-court kick-out, so that courts, not the executive, decide whether a whistleblower retaliation claim moves forward, foreclosing pocket-veto-by-delay.
  2. Revival of the Office of Technology Assessment (OTA), restoring the legislative-branch technical-assessment capacity Congress eliminated in 1995, so that policymaking does not depend solely on executive-branch evaluations.
  3. Federally funded technical assistance for state attorneys general, anchored at the FTC's Office of Technology, to distribute enforcement capacity across multiple jurisdictions so that an executive-branch retreat does not collapse all enforcement.

(Source: asteriskmag.substack.com)

The essay advances an institutional-design counter-position to the procurement-driven and pre-release-vetting frames the author describes as dominant in US AI policy. It characterizes both as brittle relative to political risk and argues for structural redundancy as the binding governance design principle.

Relationships

Sources

Umbrella page; sources live on the linked pages. Created 2026-05-11 during the v4.0 backlog-close pass.