Representative Pramila Jayapal has published a policy framework proposing federal charters as the mandatory condition for operating in the U.S. AI industry — with a structural separation between compute and models at its core.
What Happened
On October 1, 2026, Representative Pramila Jayapal (WA-07) published a document she calls the National AI Charter Act framework. The release was dateline Seattle. Her own quoted language refers to “the crux of this bill” and “this legislation.” What is publicly available is the framework outline — no bill number, no statutory text, no cosponsor list, no committee referral, and no scheduled vote. Nothing has been enacted.
The framework proposes that every AI company operating in the United States must hold a federal charter to do so. Companies valued at one billion dollars or more would require Congressional approval to receive that charter. Companies below that threshold would apply to a newly created body called the National AI Charter Authority. Every charter, regardless of how it was granted, would require reapproval every ten years.
The framework also proposes a Monopoly Data Tax whose proceeds would flow into an AI Public Fund. That fund would distribute money to Americans through instruments it calls AI Bonds. Union recognition would be a condition of holding a ten-year charter. The document states that the new authorities would not preempt or replace existing federal, state, and local laws.
The key insight: The lead mechanism here is not a fine and not a tax. It is a revocable permission to operate. The framework describes charter requirements as “a mandatory price of entry into the AI industry” and the condition for remaining in it. A penalty can be estimated and priced in. A permission that can be withdrawn cannot.
The Structural Read
The most structurally significant sentence in the framework document is not about charters or taxes. It is one sentence about ownership.
The framework states that chartered companies can own either AI infrastructure or AI systems, but never both. The document describes this as preventing chip and cloud giants from controlling dependent models.
That single clause, if ever enacted, would separate the compute layer from the model layer by law. The arrangement it describes is one where the same organization supplies the underlying compute and also builds the models that run on it. The framework names that arrangement as the thing the clause is meant to prevent.
The framework goes further. It would bar chartered companies from owning or controlling non-AI businesses in sectors such as banking, retail, and pharmaceuticals. An AI company holding a charter could not also operate in those adjacent verticals.
Lina Khan — quoted in the framework release
“For generations we have required banks, drug makers, and nuclear operators to meet public terms before they do business, and AI companies should be no different.”
The framework names that lineage explicitly. It cites the regulatory model for national banks, with round-the-clock oversight and structural separation between banking and commerce. It also cites the federal testing and approval requirements that apply to nuclear materials, hazardous chemicals, and dangerous weapons.
Amba Kak of AI Now points to the same lineage, describing it as governing banks, utilities, and critical infrastructure. The framework’s design borrows the logic of those regimes: pre-authorization before operation, not liability after harm.
The quarterly compliance mechanism extends that logic inward. Quarterly reports certifying safety and charter compliance would be signed by individual executives under penalty of perjury. The obligation falls on named people, not only on legal entities.
Three Implications of the Framework’s Design
IMPLICATION 1 — THE OWNERSHIP SEPARATION If enacted, the “never both” clause would require any organization currently owning AI infrastructure and AI systems to choose one or divest the other. The framework does not describe a transition mechanism, a grace period, or a valuation method for that divestiture. Those details do not exist in the published document.
IMPLICATION 2 — THE REVOCABLE PERMISSION LOGIC The framework calls charter requirements “a mandatory price of entry” and the condition for remaining in the industry. A ten-year reapproval cycle means this permission is not permanent. Companies would need to demonstrate ongoing compliance to win it again — not merely pay a penalty and continue. That changes how compliance is weighted against other strategic decisions.
IMPLICATION 3 — THE DISTANCE FROM LAW A framework document is not a bill. A bill is not a law. None of these proposals have statutory text, a bill number, a committee referral, or a scheduled vote. Endorsements from Lina Khan, Jonathan Kanter, Tim Wu, Randi Weingarten, and Lisa Gilbert of Public Citizen are statements of public support. They are not cosponsorships. The framework’s proposals describe an intent; the legislative process would determine whether any of them ever become operative.
The Bottom Line
The National AI Charter Act framework published by Representative Jayapal on October 1, 2026 is a policy document, not a law, not a bill, and not statutory text. Its sharpest structural proposal — that a chartered company may own AI infrastructure or AI systems, but never both — is addressed, in the document’s own words, to preventing chip and cloud giants from controlling dependent models.
The charter-as-permission design, borrowed explicitly from national banking and high-risk materials regulation, treats the right to operate as temporary and conditional rather than as a default. How far that design travels from a framework outline to enacted law is a separate question, and one the document itself does not answer.
Nothing in this article is legal advice and nothing is investment advice.
Sources: National AI Charter Act Framework (PDF) — jayapal.house.gov · Press Release — Rep. Pramila Jayapal, 1 October 2026
91,000+ executives read Business Engineer for the AI strategy frameworks cited by ChatGPT, Claude, and Perplexity.
Every quotation and figure above comes from two documents published by Rep. Pramila Jayapal on 1 October 2026: the press release at jayapal.house.gov and the National AI Charter Act Framework outline linked from it. Both were read directly, and the framework PDF was fetched from house.gov and checked byte-for-byte against a second copy. What was published is a policy framework and a press release. There is no bill number, no statutory text, no cosponsor list, no committee referral and no scheduled vote, and nothing has been enacted.
The documents are not fully consistent on this point: the release describes a policy framework while the congresswoman’s quoted words refer to “this bill” and “this legislation”. Both are reported above as they appear. Nothing above takes any position on whether the framework is desirable, workable, necessary or likely to advance, and nothing above predicts any legislative, political, legal or industry outcome. The $1 billion valuation threshold, the ten-year reapproval term and the separation of AI infrastructure from AI systems are quoted from the framework outline; the documents do not say how a valuation would be measured or by whom.
No AI company is named above because neither document names one. Nothing above describes what banking, nuclear, chemical or weapons regulation actually requires beyond the comparison the documents themselves draw, because those regimes were not examined here. The endorsements quoted are statements of support. They are not votes, not cosponsorships and not evidence about whether anything will advance. Also absent, because the documents do not give them: any hearing, cost estimate or effective date. Nothing here is legal advice and nothing here is investment advice.
