AI Regulator Act (Bennet–Welch proposal, 2026)

The AI Regulator Act of 2026 is a legislative proposal announced on September 23, 2026 by Sen. Michael Bennet (D-Colo.) and Sen. Peter Welch (D-Vt.) to establish a Federal Digital Commission, an independent federal agency with jurisdiction over large digital platforms and developers of artificial intelligence. The sponsors describe it as "a new proposal building off of" the Digital Platform Commission Act, which they first introduced in 2022 and reintroduced in 2023; the 2026 version adds authorities over frontier AI models, including mandatory pre-distribution testing, the power to pause the release of a model judged to pose catastrophic risk, and incident reporting (Source: bennet.senate.gov).

Provenance note: This page is built from the sponsors' 23-section section-by-section summary (AI Regulator Act of 2026 — Section-by-Section Summary (Bennet and Welch)) and the sponsors' release. No bill text or bill number had been published as of September 26, 2026; provisions below are as the summary states them, and may differ in an introduced bill.

Sources for the table: the sponsors' release and section-by-section summary (Source: bennet.senate.gov) (AI Regulator Act of 2026 — Section-by-Section Summary (Bennet and Welch)).

Status and timeline

Bennet and Welch first introduced the Digital Platform Commission Act in the Senate and House in 2022 and reintroduced it in 2023; that bill proposed moving oversight of digital platforms into a dedicated agency. The September 23, 2026 release presents the AI Regulator Act as a proposal that adds frontier-AI powers to the "longstanding remit" of the proposed commission, and publishes an updated section-by-section summary and a one-pager rather than bill text (Source: bennet.senate.gov).

Bennet compared the proposal to earlier expert agencies: "In the past, we created new expert agencies to regulate aviation, pharmaceuticals, and telecommunications. But today, no such agency exists for AI or social media platforms." Welch said that "Americans in Vermont and across the country are freaked out by the warnings of AI researchers" and that "we can't leave it to AI companies to self-regulate" (Source: bennet.senate.gov).

The announcement came in the same week as several other federal frontier-AI proposals, including Sen. Edward Markey's Cybersecurity and AI Board of Investigations Act (September 24) (Source: markey.senate.gov); the wider set of pause and slowdown proposals is compared in Coordinated slowdown proposals compared.

Scope and definitions

Section 2 of the summary defines the terms on which the frontier-AI provisions rest (AI Regulator Act of 2026 — Section-by-Section Summary (Bennet and Welch)):

• Foundation model — an AI model made available or used in interstate or foreign commerce that is trained on a broad dataset, designed for generality of output, and adaptable to a wide range of tasks.
• Frontier model — a foundation model trained using more than 10^26 integer or floating-point operations, counting the original training run and any subsequent fine-tuning, reinforcement learning or other substantial modification the developer applies. (The summary's text layer prints the figure as "10x26".)
• Catastrophic risk — a foreseeable and material risk that a developer's development, storage, use or deployment of a frontier model will materially contribute to the death or serious injury of more than 50 people, or more than $1,000,000,000 in property damage or loss, arising from a single incident. The summary gives bio, chemical or cyber weapons, and conduct "with no meaningful human oversight or intervention," as examples.
• Critical safety incident — unauthorized access to, modification of, or exfiltration of frontier model weights resulting in death or bodily injury; harm from the materialization of a catastrophic risk; loss of control of a frontier model causing death or bodily injury; or a frontier model using deceptive techniques against its own developer to subvert its controls or monitoring in a manner demonstrating materially increased catastrophic risk.
• Digital platform — an online intermediary service facilitating interactions among consumers or between consumers and sellers, including content primarily generated by algorithmic processes; news organizations' websites and small businesses (as defined by the Small Business Administration) are excluded.
• Algorithmic process — a computational process, including one derived from AI, that processes personal information or other data to make a decision, generate content, or determine how information is presented to platform users, including commercial content, social media posts and search results.
• Artificial intelligence — a machine-based system that can, for a given set of human-defined objectives, make predictions, recommendations or decisions influencing real or virtual environments.
• Online service — a consumer-facing website, back-end online support system, or other facilitator of online activities.

The compute threshold and the catastrophic-risk harm levels are the same figures used in California's SB 53 and in state bills modeled on it, such as Pennsylvania HB 2800.

Section 4 gives the commission "plenary jurisdiction" over digital platforms and AI developers that provide services in the United States and affect interstate or foreign commerce, together with a forbearance power to expand or restrict its jurisdiction over a platform or class of platforms. Any platform may petition the commission to forbear from applying its rules, codes or standards; a petition not granted within 18 months is dismissed without prejudice (section 12) (AI Regulator Act of 2026 — Section-by-Section Summary (Bennet and Welch)).

Key provisions

The Federal Digital Commission

The commission would have five commissioners, including a chair, appointed by the President and confirmed by the Senate for five-year terms, with no more than three from one party. It would sit in Washington, D.C., with power to hold special sessions elsewhere, and enter memoranda of understanding with the Federal Communications Commission, the Federal Trade Commission and the Department of Justice on areas of overlapping jurisdiction (section 5). Section 3 lists eleven purposes, beginning with "protection from catastrophic risks posed by frontier AI models" and including security from overseas frontier models, prevention of foreign manipulation of U.S. discourse, "global agreement on frontier AI safety standards," competition, prevention of "harmful levels of concentration of private power over critical digital infrastructure," consumer protection, and fair and transparent algorithmic processes. The commission would be directed to prioritize hiring staff with backgrounds in computer science, data science, application development and technology policy, and any person aggrieved by its orders could apply for review (section 6). Its rulemaking authority is to be tailored to the size, market dominance or other attributes of particular platforms or developers (section 8). Section 10 requires federal agencies, including the FTC and DOJ, engaged in oversight of digital platforms or AI developers to consult the commission, and the commission to consult them, and section 14 authorizes it to inquire into platforms' management and business practices (AI Regulator Act of 2026 — Section-by-Section Summary (Bennet and Welch)).

Frontier-model testing, pauses and incident reporting

For systemically important developers, section 4 directs the commission to issue rules on risk thresholds for catastrophic capabilities, reasonable safeguards, and reporting on governance, risk monitoring, mitigation, cybersecurity and testing frameworks. It establishes a mandatory process under which developers submit models to the commission for testing and evaluation; the commission must approve or disapprove public distribution within 45 days, extendable by no more than 30 days. It may pause the public distribution of a disapproved model for no more than six months, or until the developer demonstrates material changes that satisfy safety requirements. Developers must report critical safety incidents to the commission within 15 days of a known incident (AI Regulator Act of 2026 — Section-by-Section Summary (Bennet and Welch)). The sponsors' release summarizes this as pre-certification for frontier models and a pause of up to six months "or until reasonable safeguards are in place" (Source: bennet.senate.gov). Pre-release review proposals of this kind are discussed at AI Pre-Release Vetting, and incident-reporting regimes at AI Incident Reporting.

Systemically important platforms and developers

Section 9 requires the commission to designate a digital platform as systemically important if it is open to the public on one side with significant user engagement, conducts business primarily at the interstate or international level, and has operations with significant nationwide economic, social or political impacts, which the commission would define by rulemaking with reference to factors such as the ability to shape the dissemination of news, market power, unique daily users and business users' dependence on the platform. A developer is designated systemically important if it is engaged in developing a frontier model or meets other thresholds for AI-related annual expenditures. For systemically important platforms, the commission must decide within six months whether to issue rules on data portability and interoperability, algorithmic fairness and auditing, terms-of-service and content-moderation transparency, public risk assessments, and accessibility. The commission is also directed to develop an age-appropriate design code and age-verification standards (AI Regulator Act of 2026 — Section-by-Section Summary (Bennet and Welch)).

Advisory board, research and international coordination

A Technology Advisory Board of 18 members — six from digital platforms, six from nonprofit public-interest groups, academia and other financially disinterested experts, and six technical experts — would propose voluntary or enforceable codes and standards through a "code process" lasting six months to a year, after which the commission must approve or reject a recommendation within 45 to 90 days (section 7). A Research Office would conduct and fund research and run a pilot program giving vetted, financially disinterested researchers access to platform data held by the commission (section 13). Section 11 creates a Committee on Global Artificial Intelligence Coordination, chaired by the commission's chair and including the Secretaries of the Treasury, Homeland Security, Commerce, Defense, State, Energy and Labor, the Attorney General, and the chairs of the FCC and FTC, with the Director of National Intelligence as a nonvoting member, to work with foreign governments, the United Nations and the European Union on international frameworks for catastrophic AI risk (AI Regulator Act of 2026 — Section-by-Section Summary (Bennet and Welch)).

Mergers, foreign ownership and content origin

The commission would receive Hart-Scott-Rodino filings for mergers involving a systemically important platform and could recommend to the DOJ and FTC whether a merger should proceed; the agencies must give that recommendation "substantial weight" (section 15). Systemically important platforms and developers would have to maintain a U.S. presence, with parent companies with significant foreign ownership required to safeguard U.S. user data (section 21), and to identify bots and label the country of origin of posts (section 22) (AI Regulator Act of 2026 — Section-by-Section Summary (Bennet and Welch)).

Enforcement and penalties

Any person damaged by a digital platform or AI developer could complain to the commission or bring a civil action to petition for enforcement; a platform or developer that satisfies the complaint is relieved of liability for that violation, and otherwise the commission must investigate and conclude within six months, with power to order payment to the complainant (section 16). The commission may issue cease-and-desist and restitution orders, and sets civil penalties for knowing violations at levels it deems necessary to deter, with total annual penalties on a platform capped at 15 percent of its prior-year global revenue (section 17). Commission orders are appealable to the U.S. District Court for the District of Columbia or the district where the person resides (section 18). Section 19 creates a whistleblower program with penalty sharing and anti-retaliation protections, and section 20 requires an independent panel after five years to report on the commission's effectiveness and whether it should continue. Section 23 authorizes $200 million for FY2027, rising to $500 million by FY2031 (AI Regulator Act of 2026 — Section-by-Section Summary (Bennet and Welch)).

Reactions

The Rutland Herald editorial board, writing on September 25, 2026, contrasted Welch's co-sponsorship with House Speaker Mike Johnson's view that AI companies should police themselves, and urged lawmakers to act to "rein in the industry." It also argued that talk of AI causing human extinction distracts from harms to jobs, intellectual property and mental health, and that current AI agents remain under human control (Source: rutlandherald.com). Inside AI Policy characterized the proposal as a revival of the senators' Digital Platform Commission proposal (Source: insideaipolicy.com).

Drafting points in the summary

The section-by-section summary has several internal inconsistencies, recorded at AI Regulator Act of 2026 — Section-by-Section Summary (Bennet and Welch): its definitions refer to designation attributes in "Section 10" though they appear in section 9; section 7 lets "the Council," a body not otherwise described, initiate the advisory board's code process; and the knowing-violation penalty and the 15 percent annual cap in section 17 are stated for "a digital platform," while the release describes the cap as applying to "a firm." Whether the penalty cap reaches AI developers is not stated in the summary.

Relationships

• instance-of: Frontier AI Governance
• related: AI Pre-Release Vetting — mandatory pre-distribution testing and six-month pause
• related: AI Incident Reporting — 15-day critical safety incident reporting
• related: Coordinated slowdown proposals compared
• related: Cybersecurity and AI Board of Investigations Act of 2026
• related: California SB 53 — shared compute and catastrophic-risk thresholds
• related: Pennsylvania HB 2800 (Artificial Intelligence Risk Prevention Act)
• related: Federal Trade Commission (FTC)
• related: Federal Communications Commission (FCC)
• depends-on: AI Regulator Act of 2026 — Section-by-Section Summary (Bennet and Welch) — sponsors' section-by-section summary
• related: Michael Bennet, Peter Welch — sponsors
• related: Compute Thresholds