# Artificial Intelligence Policy — Federal AI Governance & Regulation

Federal regulation of artificial intelligence AI is evolving rapidly as AI systems — including large language models, autonomous vehicles, facial recognition, algorithmic decision-making, and generati

- Canonical: https://policyrisk.com/wiki/artificial-intelligence-policy
- As of: 2026-11-18
- Plain-text twin of the page above; the HTML and this file carry the same facts.

## Facts
- Domain: Technology
- Category: Emerging Technology
- Page type: topic
- U.S. Code anchor: none
- Last verified: not recorded

## Article

Federal regulation of **artificial intelligence (AI)** is evolving rapidly as AI systems — including large language models, autonomous vehicles, facial recognition, algorithmic decision-making, and generative AI — transform virtually every sector of the economy and government. Unlike many technology areas, the U.S. has not enacted a **comprehensive federal AI statute** as of 2026. Instead, AI governance operates through a patchwork of **executive orders**, **agency-specific regulations**, **voluntary frameworks**, and **existing laws** applied to AI contexts. The most significant federal actions include: the **National AI Initiative Act of 2020** (15 U.S.C. §§ 9401–9461, enacted as part of the NDAA), which established the **National AI Initiative Office** and directed AI research coordination across agencies; President Biden's **Executive Order 14110** (October 2023) on "Safe, Secure, and Trustworthy AI," which imposed reporting requirements on developers of powerful AI systems, directed agencies to address AI risks, and established AI safety testing standards; and the **NIST AI Risk Management Framework** (AI RMF, published 2023), a voluntary governance framework for managing AI risks. Individual agencies regulate AI within their existing jurisdictions: the **[FDA](fda-drug-approval.md)** governs AI in medical devices, the **[FTC](federal-trade-commission.md)** addresses deceptive or unfair AI practices, the **EEOC** addresses AI-driven employment discrimination, the **DOT/NHTSA** regulates autonomous vehicles, and the **financial regulators** address AI in credit decisions and trading. The absence of a comprehensive federal AI law has left states to fill gaps — with states like Colorado, California, and others enacting AI-specific legislation. See [Data Privacy Law](data-privacy-law.md) for the related privacy framework and [NIST Standards & Technology](nist-standards-technology.md) for the AI Risk Management Framework's home agency.

## Current Law (2026)

| Parameter | Value |
|-----------|-------|
| Comprehensive federal AI statute | None as of 2026 — governance through executive orders, agency rules, and existing law |
| National AI Initiative Act | 15 U.S.C. §§ 9401–9461 (2020) — research coordination, National AI Initiative Office |
| Key executive order | EO 14110 (Biden, October 2023) — **rescinded** January 2025; Trump EO "Removing Barriers to American Leadership in Artificial Intelligence" (January 2025) replaced it |
| NIST AI RMF | Voluntary AI Risk Management Framework (published January 2023) |
| Agency jurisdiction | FDA (medical AI), FTC (consumer protection), EEOC (employment), DOT (autonomous vehicles), financial regulators (credit/trading) |
| State action | Colorado AI Act (2024), California AI transparency bills, Illinois BIPA (biometric AI) |
| International context | EU AI Act (2024) — the world's first comprehensive AI regulation; U.S. takes a more sectoral approach |

## Legal Authority

- **15 U.S.C. §§ 9401–9461** — National Artificial Intelligence Initiative Act of 2020
- **Executive Order 14110** (October 30, 2023) — Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence
- **NIST AI Risk Management Framework** (AI RMF 1.0, January 2023) — voluntary framework for AI governance
- Existing statutes applied to AI: **FTC Act** (§ 5, unfair/deceptive practices), **Civil Rights Act** (Title VII, employment discrimination), **ECOA** (fair lending), **FDA** (medical device regulation), **ADA** (disability discrimination)

## How It Works

Rather than enacting a single comprehensive AI statute (as the EU did with the AI Act), the U.S. applies **existing agency authorities** to AI within each sector. The [FTC](federal-trade-commission.md) uses Section 5 authority over unfair or deceptive practices to police misleading AI claims and harmful outputs. The [EEOC](eeoc-employment-discrimination.md) has issued guidance that AI hiring tools are subject to Title VII disparate impact analysis — discriminatory outcomes matter regardless of intent. The FDA regulates AI-enabled medical devices (diagnostic algorithms, imaging analysis, clinical decision support) through existing premarket pathways — over **900 AI/ML-enabled medical devices** have been authorized. Financial regulators (OCC, CFPB, Federal Reserve) address AI in credit decisions; NHTSA develops AV safety standards. The most comprehensive executive action was **Executive Order 14110** (October 2023), which directed reporting requirements for developers of AI models trained above certain compute thresholds, mandated red-teaming safety testing, directed every major agency to appoint Chief AI Officers and address AI risks in their sectors, and required protections against AI-enabled fraud and algorithmic discrimination — but executive orders are fragile, and EO 14110 was revoked in early 2025.

The **NIST AI Risk Management Framework** (published January 2023) fills part of the policy vacuum with a voluntary, flexible governance structure organized into four functions: **Govern** (establish AI oversight structures), **Map** (identify risks), **Measure** (assess and track), and **Manage** (prioritize and respond). It is not legally binding but widely adopted by industry and government and increasingly used as a benchmark by regulators. In the absence of federal legislation, states have created a compliance patchwork: **Colorado's AI Act** (2024) requires developers and deployers of high-risk AI to avoid algorithmic discrimination and provide transparency; **Illinois' BIPA** has generated thousands of lawsuits over facial recognition; **California** has enacted AI transparency requirements; and dozens of states are considering legislation. This fragmentation creates compliance burdens for multi-state businesses and has become the strongest political argument for federal preemptive AI legislation.

## How It Affects You

**If you're a consumer affected by AI in credit, hiring, insurance, or medical decisions**: You have more legal rights than you may realize — but enforcing them requires knowing which law applies. For **credit decisions**: the Equal Credit Opportunity Act (ECOA) and Fair Housing Act prohibit AI models that produce discriminatory outcomes, regardless of intent — the CFPB has confirmed that disparate impact claims apply to algorithmic underwriting. If you're denied credit, you're entitled to an **adverse action notice** (Reg B) explaining the reasons. For **hiring**: the EEOC's 2023 guidance on AI hiring tools affirms that automated screening and assessment tools must comply with Title VII — disparate impact analysis applies. New York City's Local Law 144 (effective 2023) requires employers using automated employment decision tools to conduct annual bias audits and notify candidates. For **health insurance**: the ACA prohibits algorithmic discrimination based on pre-existing conditions, and CMS has issued guidance against AI-driven prior authorization denials that don't meet medical necessity standards. Practical step: if you believe an AI-driven decision was discriminatory, file a complaint with the relevant agency (CFPB for credit, EEOC for employment, HHS OCR for healthcare).

**If you're a business deploying AI systems**: Your existing legal obligations under sector-specific law apply fully to AI-driven decisions — "the algorithm made the decision" is not a defense. The FTC has made clear under Section 5 (unfair or deceptive practices) that companies are responsible for discriminatory, deceptive, or harmful AI outputs. The EEOC has issued guidance that AI hiring tools are subject to Title VII disparate impact analysis — if your vendor's tool produces adverse selection rates for protected groups, you may have liability even without discriminatory intent. For **high-risk AI systems**: the NIST AI Risk Management Framework (AI RMF, January 2023) provides voluntary governance guidance — documenting your risk assessment, testing for bias, maintaining human oversight, and transparency with users are the recommended practices that regulators will likely benchmark enforcement against. **State laws are moving faster than federal**: the Colorado AI Act (effective 2026 for high-risk AI in consequential decisions), Illinois AI Video Interview Act, and New York City Local Law 144 create specific obligations. Track state legislation actively — patchwork compliance is the near-term reality.

**If you're developing or training AI models at scale**: The Trump administration rescinded Biden's EO 14110 (AI safety reporting requirements) in January 2025, removing the mandatory safety reporting obligations for large-scale model developers. However, sector-specific regulations still apply to AI deployed in regulated industries (FDA for medical AI, OCC/CFPB for financial AI, FCC for communications). For models with potential national security implications: the Commerce Department's Bureau of Industry and Security (BIS) has issued export controls on advanced AI chips — see [NIST Cybersecurity Framework](nist-cybersecurity-framework.md) for the broader technology standards context — and has authority under the AI Diffusion Rule to regulate training of frontier models. NIST's AI Safety Institute (created by EO 14110, now reorganized under the Trump administration as the AI Safety Institute Consortium) continues voluntary AI safety evaluations. For frontier model developers: voluntary commitments made to the White House in 2023 (on red-teaming, security, safety reporting) remain in place for the original signatories but are not legally binding.

**If you're a worker whose job is affected by AI automation or AI-powered management**: AI-driven automation is reshaping labor markets across industries — but its deployment in workplace management (productivity monitoring, scheduling algorithms, performance scoring) creates immediate legal issues distinct from long-term displacement. Workplace surveillance AI that monitors keystrokes, location, or communications must comply with state privacy laws — California, New York, and Connecticut have specific notice requirements for electronic monitoring. AI performance management systems that use output metrics to make adverse employment decisions must comply with employment discrimination law; if the metrics systematically disadvantage workers in protected categories, the employer faces disparate impact liability. For collective bargaining: the NLRB has issued guidance that employers must bargain with unions over the introduction of AI systems that affect working conditions (including monitoring and performance management systems). If you're in a union: raise AI deployment in bargaining. If you're not: document any AI-driven adverse employment actions and consult an employment attorney — the legal framework is still developing but existing anti-discrimination law applies now.

## State Variations

AI regulation varies significantly by state and is rapidly evolving:

- Colorado AI Act (2024) — first comprehensive state AI anti-discrimination law
- Illinois BIPA — biometric privacy law generating significant AI litigation
- California — multiple AI transparency and safety bills
- New York City Local Law 144 — requires bias audits for automated employment decision tools
- Most states have no AI-specific legislation — existing consumer protection, privacy, and anti-discrimination laws apply

## Implementing Regulations

Comprehensive federal AI regulation is still emerging; OMB memorandum M-24-10 and the AI Executive Order framework guide agency-specific implementation. Key existing regulations applied to AI contexts include:

- **15 CFR Part 742** — BIS export controls on AI-related technologies, including dual-use AI chips and advanced semiconductors, requiring export licenses for certain high-performance computing hardware to restricted destinations
- **48 CFR Part 204** — Federal acquisition of AI systems, incorporating DFARS and FAR AI procurement requirements that agencies must follow when acquiring AI-enabled products and services

## Pending Legislation

Multiple AI bills have been introduced in the 119th Congress. See [Technology Regulation](tech-regulation.md) for broader legislative activity on AI governance, algorithmic accountability, and AI in federal procurement.

Senate Banking Chair Tim Scott (R-S.C.) and Sen. Bill Hagerty (R-Tenn.) introduced the ICTS Supply Chain Security Act, which would give Commerce new powers to block technology from China, Russia, North Korea, Cuba and Iran from critical systems and supply chains for artificial intelligence. The bill would also create a new Senate-confirmed Assistant Secretary at Commerce to oversee information and technology supply chains and codify the Office of Information and Communications Technology and Services within the Bureau of Industry and Security (POLITICO's Morning Money, June 29, 2026).

## Recent Developments

- **Trump administration reversal on AI safety (January 2025)**: President Trump rescinded Biden's Executive Order 14110 on his first day in office and issued his own EO: "Removing Barriers to American Leadership in Artificial Intelligence." The new order directs agencies to revoke or revise any Biden-era AI policies inconsistent with a pro-innovation, deregulatory approach and tasks the Office of Science and Technology Policy (OSTP) with developing a new national AI action plan. The mandatory safety reporting requirements for large-scale AI developers — the most contentious provision of EO 14110 — were eliminated.
- **NIST AI Safety Institute restructured**: The AI Safety Institute (AISI), created within NIST by EO 14110, was reorganized and rebranded under the Trump administration. Its focus shifted toward U.S. competitiveness and evaluating AI systems for government use, rather than the safety-testing emphasis of the Biden era. The institute's relationships with international partners (UK, EU AI safety institutes) were downgraded.
- **AI chip export controls**: The Biden administration's "AI Diffusion Rule" — restricting exports of advanced AI chips (Nvidia H100/H200 class) to most of the world except close allies — was issued in January 2025 in the final days of the administration. The Trump administration subsequently reviewed and revised the rule, with ongoing debate over whether restrictions impede U.S. AI competitiveness or legitimately protect national security.
- **State AI laws going into effect**: Colorado's AI Act (2024), the first comprehensive state AI anti-discrimination law, applies to high-risk AI systems in consequential decisions (credit, insurance, employment, housing) as of 2026. California enacted additional AI transparency requirements. The patchwork of state laws is intensifying industry pressure for federal preemption through a comprehensive federal AI statute — but none has passed in the 119th Congress.
- **EU AI Act implementation**: The EU's AI Act began applying its first provisions in 2025, banning certain unacceptable-risk AI practices (mass biometric surveillance, social scoring). High-risk AI system requirements apply from 2026. U.S. companies operating in the EU face compliance obligations even as U.S. domestic requirements remain lighter — creating a de facto compliance floor for multinational AI developers.

**FINRA issues observations on AI agents in financial services (January 2026)**: FINRA published a regulatory blog post documenting its observations of agentic AI adoption among broker-dealer member firms, identifying six categories of AI agents in early deployment: conversational agents, software development agents, fraud detection agents, trade and AML surveillance agents, process automation agents, and trade execution agents. FINRA flagged specific risks including autonomous action without human validation, agents exceeding intended scope of authority, auditability gaps in multi-step reasoning, and data sensitivity concerns. The publication does not create new legal requirements but signals that FINRA's existing technology-neutral rules and federal securities laws apply to agentic AI deployments, and it encourages firms to proactively engage with FINRA as they develop agentic AI strategies. [FINRA, *Emerging Trend in GenAI: Observations on AI Agents* (Jan. 27, 2026)](https://www.finra.org/media-center/blog/observations-on-ai-agents)

An interim final rule titled "AI Action Plan Implementation" (RIN 0694-AJ90), submitted by agency code 0694 (Commerce Department/NIST) to OIRA on February 26, 2026, was withdrawn by OIRA on March 13, 2026, before taking effect. The rule was classified as neither economically significant nor major. The withdrawal indicates that at least one formal regulatory vehicle for implementing the Trump administration's AI Action Plan — directed by the January 2025 executive order — did not advance, though the underlying action plan effort continues. [Reginfo.gov, OIRA Withdrawn Rule 0694-AJ90: AI Action Plan Implementation (Mar. 13, 2026)](https://www.reginfo.gov/public/Forward?SearchTarget=RegReview&textfield=0694-AJ90)

**DOL launches 'Make America AI-Ready' initiative (March 2026)**: The U.S. Department of Labor announced the "Make America AI-Ready" initiative on March 24, 2026, signaling a federal workforce-focused effort to prepare workers and employers for AI-driven changes in the labor market. The initiative reflects the Trump administration's broader pro-innovation AI posture and extends federal AI policy activity into workforce development and training. Further details on the initiative's specific programs, funding, or legal authorities were not available from the announcement alone. [U.S. Department of Labor, *Make America AI-Ready* (Mar. 24, 2026)](https://www.dol.gov/newsroom/releases/osec/osec20260324)

**Florida AG investigates OpenAI (April 2026)**: Florida Attorney General James Uthmeier announced on April 9, 2026, that his office is opening an investigation into OpenAI and its ChatGPT chatbot. Uthmeier cited concerns about OpenAI data and technologies potentially falling into the hands of the Chinese Communist Party, as well as ChatGPT's alleged links to criminal behavior including child sexual abuse material. The investigation represents a notable instance of state-level law enforcement scrutiny of a major AI developer, adding to the patchwork of state-level AI oversight activity alongside legislative efforts in Colorado, California, and other states. [*Florida attorney general launches investigation into OpenAI, ChatGPT*, The Hill (Apr. 9, 2026)](https://thehill.com/policy/technology/5824392-florida-investigates-openai-chatgpt/)

The Florida Attorney General's investigation into OpenAI escalated on April 21, 2026, when AG James Uthmeier announced a formal criminal investigation and served OpenAI with subpoenas. Uthmeier linked the probe in part to the individual accused of fatally shooting two people at Florida State University, alleging a connection to ChatGPT. This marks a significant escalation from the initial civil inquiry announced April 9, 2026, and represents one of the first instances of a state attorney general pursuing a criminal investigation targeting a major AI developer. [*Florida attorney general issues subpoenas in OpenAI criminal probe*, The Hill (Apr. 21, 2026)](https://thehill.com/policy/technology/5841734-openai-florida-state-shooting/)

A CRS In Focus report (published Oct. 14, 2026) states that on February 27, 2026, President Trump directed federal agencies to stop using Anthropic's technology, and Secretary of Defense Hegseth directed DOD to designate Anthropic a supply-chain risk to national security. The Department of Commerce imposed export controls on Anthropic's Mythos 5 and Fable 5 models on June 12, 2026, and lifted them on June 30, 2026. On August 27, 2026, a federal judge vacated and permanently enjoined the Section 3252 designation, and on September 25, 2026, the D.C. Circuit upheld DOD's Section 4713 designation. [CRS, IF13217: Federal Government and Anthropic](https://www.congress.gov/crs-report/IF13217)

A CRS In Focus report published June 10, 2026 states that FDA published final guidance in August 2025 with nonbinding recommendations on information to include in predetermined change control plans for certain premarket submissions of AI-enabled devices. The report also states that FDA issued final guidance on device cybersecurity in February 2026, which in part addresses statutory requirements for cyber devices. In September 2025, FDA requested public comment, due December 2025, on measuring real-world performance of AI-enabled medical devices. [CRS, IF13245: FDA Regulation of AI-Enabled Devices](https://www.congress.gov/crs-report/IF13245)

The Five Eyes intelligence alliance (U.S., U.K., Canada, Australia and New Zealand) issued a rare joint warning that frontier AI capable of crippling governments and businesses is close, stating that "the timeline is not years, it is months." The joint bulletin calls for a "whole-of-society response" to accelerating AI-enabled cyber risk and urges boards and executives to ensure cyber resilience and use AI deliberately for defense. (Axios AM, June 24, 2026)

The House Science Committee, chaired by Brian Babin (R-Texas), began marking up measures on topics including Commerce Department risk assessments of powerful AI models, content authentication, and flaw reporting (Punchbowl News, June 25, 2026). The newsletter reports this markup as taking place the morning of its June 25, 2026 edition.

OpenAI began a limited release of GPT-5.6 on June 26, 2026, with access restricted at the behest of the U.S. government to around 20 companies whose participation the government approved (Axios Closer, June 26, 2026). The newsletter reports that this follows similar U.S. restrictions on Anthropic's Fable 5 and Mythos 5 models (Axios Closer, June 26, 2026). OpenAI said it expects to expand access to more companies the following week (Axios Closer, June 26, 2026).

The Trump administration requested that OpenAI limit the rollout of its GPT-5.6 model, citing national security concerns, and OpenAI agreed to restrict the release (Axios Alerts, June 26, 2026). The alert reports that the restriction followed the government's request.

The Commerce Department allowed Anthropic to restore access to its Mythos 5 cybersecurity model for a limited number of trusted users, which the newsletter describes as a sign of de-escalation in the administration's standoff with Anthropic. Access to Anthropic's Fable 5 model, offline since June 12, had not been restored as of the report; the Pentagon and National Security Agency still had to approve it. (Axios AM, June 27, 2026)

On July 1, 2026, the Federal Trade Commission sought public comment on a proposed policy statement addressing concerns that AI companies may manipulate their AI systems' outputs contrary to reasonable consumer expectations for objectivity and accuracy. The proposed statement says that distorting AI outputs to achieve undisclosed ideological objectives could violate Section 5 of the FTC Act, and that a state law such as Colorado's Artificial Intelligence Act is "impliedly preempted to the extent it conflicts with a federal regulatory scheme." The Commission vote authorizing the Federal Register notice was 2-0, and public comments are due by July 31, 2026. [FTC, FTC Seeks Public Comment on Policy Statement Addressing AI Accuracy (July 1, 2026)](https://www.ftc.gov/news-events/news/press-releases/2026/07/ftc-seeks-public-comment-policy-statement-addressing-ai-accuracy)

The Trump administration lifted export controls on Anthropic's Claude Fable 5 model, which had been pulled offline 19 days earlier over security concerns. A letter from Commerce Secretary Howard Lutnick also covered Anthropic's Mythos 5 model, which the administration had allowed to be restored for a select group of government-approved organizations the previous week. Lutnick said his office had "worked closely with Anthropic to analyze and approve Fable 5" (Axios AM, July 1, 2026).

## Sources
- [Artificial Intelligence Policy — Federal AI Governance & Regulation on PolicyRisk](https://policyrisk.com/wiki/artificial-intelligence-policy)

## Related
- [data privacy law](https://policyrisk.com/wiki/data-privacy-law)
- [fitara federal it reform](https://policyrisk.com/wiki/fitara-federal-it-reform)
- [e government act](https://policyrisk.com/wiki/e-government-act)
- [fda drug approval](https://policyrisk.com/wiki/fda-drug-approval)
- [nist cybersecurity framework](https://policyrisk.com/wiki/nist-cybersecurity-framework)
- [eeoc employment discrimination](https://policyrisk.com/wiki/eeoc-employment-discrimination)
- [federal trade commission](https://policyrisk.com/wiki/federal-trade-commission)

Source: PRIA, https://policyrisk.com/wiki/artificial-intelligence-policy (as of 2026-11-18).
