Pramila Jayapal's National AI Charter Act Framework of 1 October 2026 Would License AI Companies Like National Banks: How Far the Banking Comparison Holds
On 25 February 1863, in the middle of a war it was financing with borrowed money, Congress passed the National Currency Act and created the Office of the Comptroller of the Currency as a bureau of the Treasury. The idea was simple and, at the time, radical: a bank that wanted to issue the new national notes would need a federal charter, and with the charter came a federal examiner. One hundred and sixty-three years later, a member of Congress has proposed applying the same instrument to artificial intelligence.
Practical takeaway. The National AI Charter Act framework of 1 October 2026 is a policy summary without bill text or a definition of "AI company". Its banking analogy holds for supervision, examination and separation from commerce. It departs from banking law on charter duration, congressional approval of individual companies and capability bans, which are the provisions that would decide its reach.
What Representative Jayapal released on 1 October 2026
Representative Pramila Jayapal, who represents Washington's 7th district, announced on 1 October 2026 that she is introducing a policy framework for a National AI Charter Act. The framework document runs to five pages of headings and short paragraphs. Its central rule is that every AI company would need a national charter to operate in the United States, with "non-negotiable structural, conduct, and safety requirements" as the condition of operating.
A new National AI Charter Authority would issue charters, run federal AI testing and enforce the terms. It would be led by a director, funded by charter fees, and advised by a bipartisan council of state attorneys general. A company valued at USD 1 billion or more could be chartered only with the approval of Congress; smaller companies would go through the Authority. Charters would have to be reapproved every ten years. Chartered companies would file quarterly reports certifying safety and charter compliance, signed by individual executives under penalty of perjury.
The release quotes support from Lina Khan, former Chair of the Federal Trade Commission, Jonathan Kanter, former Assistant Attorney General for Antitrust, Tim Wu of Columbia Law School and Randi Weingarten of the American Federation of Teachers. It does not give a bill number, and the framework summary does not say how an "AI company" would be identified, which matters more than any other open point.
Which national bank features the framework copies: an examiner, a prudential supervisor and a wall against commerce
The framework summary makes a specific claim: the charter is modeled on the regulatory framework for national banks, which operate under "round-the-clock oversight, strict safety and soundness mandates, and structural separation between banking and commerce." Each part of that sentence has a counterpart in the U.S. Code, in national-bank charter law and in the Bank Holding Company Act, and testing the claim against those statutes shows how close the borrowing is.
Supervision first. Under 12 U.S.C. 481, the Comptroller appoints examiners who examine every national bank "as often as the Comptroller of the Currency shall deem necessary," with power to make "a thorough examination of all the affairs of the bank," to administer oaths and to question officers under oath. The AI framework's federal inspectors "with immediate access" and its unannounced red-teaming of live systems are recognizably the same idea, applied to software that changes between examinations far faster than a loan book does.
Separation second. The Bank Holding Company Act, at 12 U.S.C. 1843(a), bars a bank holding company, with exceptions, from acquiring voting shares of any company that is not a bank. The AI framework goes a step beyond that line. A chartered company could own AI infrastructure or AI systems, but never both, could not take part in search or social media, and could not own non-AI businesses in sectors such as banking, retail and pharmaceuticals. It would also ban shared processing infrastructure, interlocking boards and cross-equity stakes. In banking terms, this combines the separation of banking from commerce with a vertical split that the banking provisions cited here do not impose: it would place the chip and cloud layer and the model layer in different hands.
Forfeiture third. Under the second power in 12 U.S.C. 24, a national bank has succession until it is dissolved by its shareholders, its franchise is forfeited for violation of law, Congress terminates it, or a receiver winds it up. The framework's "Corporate Death Penalty", which strips the charter and liquidates the assets of reckless or repeat offenders, has the same structure as that forfeiture clause.
Where the framework departs from the National Bank Act: congressional votes, ten-year charters and banned capabilities
The departures are larger than the borrowings. A national bank's charter runs until one of the four events in section 24 ends it; nothing in that list is a fixed term. The AI framework replaces that with reapproval every ten years, which would make every large AI company's existence a recurring decision. The requirement that Congress approve each charter for a company valued at USD 1 billion or more also has no analogue among national banks, whose charters are granted by the Comptroller. Applied to AI, it would turn individual licensing decisions into legislative votes, with the lobbying and delay that come with them.
The framework then adds rules that banking law does not try to write. It bans AI systems capable of evading human control, including through self-replication or resistance to shutdown, and systems that assist in creating chemical, biological, radiological or nuclear weapons. It bars systems that automate employment decisions or replace licensed professionals, skilled trades or care providers in healthcare, education and childcare. It requires chartered companies to recognize union representation and bargain to agreement as a condition of receiving and keeping the charter. Federal banking law contains no comparable list of forbidden products. These provisions read more like the licensing logic of nuclear materials than of finance, which is the second model the release names.
Why the nuclear model sits beside the banking one: testing before release and a federal kill switch
The framework's safety section would bar the public launch of high-risk AI systems without federal approval, with safety proved "inside isolated government testing facilities," under containment protocols "on par with those used for nuclear energy and defense technologies." It also requires real-time containment and kill-switch capabilities that federal safety inspectors could trigger by direct order.
California reached the kill-switch question from a different direction just over two weeks earlier. Our reading of Governor Newsom's Executive Order N-9-26 of 18 September 2026 found that it asks the Government Operations Agency and the Office of Emergency Services for recommendations by 16 November 2026 on, among other things, a kill switch for frontier models, and that it creates no duty for any developer yet. The Jayapal framework would go straight to the duty and put the trigger in the hands of a federal inspector. Between the two lies the practical question that neither text answers: what the inspector would be authorized to switch off when a model is served from many data centers and copied into customers' own systems.
What a federal AI charter would mean for patients, workers and companies that only use AI
The framework's reach extends well beyond the companies that build models. Its joint-liability provision would make AI companies and their commercial clients jointly liable for harm caused by dangerous AI applications. A hospital system that deploys a triage tool, a bank that scores applicants or an insurer that processes claims could therefore share liability with the vendor. The bans on systems that "deny benefits" or "restrict access to critical care or services" point directly at those uses.
Enforcement would run through three channels at once: the Authority, a private right of action in federal court for individuals, creators, workers and businesses, and state attorneys general with concurrent authority. The framework says it would not preempt federal, state or local law and would leave states free to enact stricter rules. For people, the most tangible proposals are a ban on training AI with data from the surveillance of locations, purchases, work routines, health records and physical features, and an AI Public Fund that would pay "AI Bonds" to all Americans from a Monopoly Data Tax and civil penalties. For patients and workers, the effect would depend on how strictly the bans on replacing licensed professionals were drafted; for citizens, on whether the charter changes what data about them can be collected at all.
How Quentir Reads It
The bank charter is a fitting precedent for one reason above the rest: it makes market entry conditional, and a conditional license changes behavior before any harm occurs. Where the framework copies federal banking law, it copies a model that has operated since 1863. Where it departs, on fixed terms, congressional approval of individual companies and outright capability bans, it leaves the current national-bank charter model behind, and those are the provisions that raise the most practical questions.
Everything still depends on the missing definition. The framework charters "AI companies" without saying what an AI company is, while the federal executive has just changed its own vocabulary: as our analysis of Executive Order 14434 of 29 September 2026 set out, federal agencies are now told to write "super intelligence" where they wrote "AI". A charter statute would need a statutory test that captures a model developer and leaves out a dental practice using transcription software. Readers who follow this thread from the federal executive orders through California's SB 53, SB 574 and Executive Order N-9-26 can read those analyses free on this site; an All-access membership adds Quentir's paid publications on the same subjects in one subscription. The question to watch when bill text appears is the definition section, since the threshold it sets decides which companies would need a charter at all.
Sources: Office of Representative Pramila Jayapal, "Jayapal Introduces Legislative Framework Establishing National Charter System to Rein in AI", 1 October 2026, and "The National AI Charter Act Framework" (framework summary, October 2026); Office of the Comptroller of the Currency, History; 12 U.S.C. 24 (corporate powers of national banks), 481 (appointment of examiners; examination) and 1843 (interests of bank holding companies in nonbanking organizations), via Cornell Legal Information Institute. All sources accessed 4 October 2026.
Published intelligence, built to inform your own decisions. Published: October 4, 2026.