Rep. Pramila Jayapal, a Democrat from Washington state, on Thursday, Oct. 1, released a plan for a National AI Charter Act that would require every AI company to hold a federal charter to do business in the U.S. A charter is a license from the government that comes with conditions a company must keep meeting to stay open. Under her plan, AI companies that don't agree, don't comply or put the public in danger would be shut down. "If you want to be an AI company in this country, you have to obtain a public charter that has a broad set of terms and conditions to operate and ensure public benefit without the harms," Jayapal said.
She modeled the idea on how the U.S. already treats national banks, which work under round-the-clock federal oversight and strict safety rules, and on the federal testing required before nuclear materials and dangerous weapons can be released. A new agency, the National AI Charter Authority, would run the system. It would be led by a director, paid for by charter fees, and advised by a bipartisan council of state attorneys general. Congress would have to approve the charter of any AI company valued at $1 billion or more, the agency would approve smaller ones, and every charter would have to be renewed every 10 years.
The plan would put the government in charge of safety at every step. Companies could not release high-risk AI to the public without federal approval, after proving it safe inside isolated government testing sites. The government and certified nonprofits could make surprise attempts to trick or break AI that is already in use, looking for hidden dangers, and federal inspectors could order a system shut off at once. AI that can escape human control or help make chemical, biological, radiological or nuclear weapons would be banned, and so would AI that replaces licensed professionals, skilled trades or human care providers in fields such as health care.
It would also reshape the industry. AI companies could not take part in search or social media, and a company could own AI systems or the chips and cloud computers that run them, but not both, which the plan says would stop chip and cloud giants from controlling the AI models that depend on them. Companies could not train AI on data gathered by tracking people, would need board seats for a safety specialist, a union representative and a public interest advocate, and would have to recognize unions. A new tax on big tech companies that built AI on stolen creative work, scraped web data or tracking would go into a public fund that pays every American directly, and executives and board members who sign false safety reports or recklessly release dangerous AI could face criminal charges.
Jayapal says the charter would not replace federal, state or local laws, and states could still pass stricter AI rules. Her release quotes about 20 supporters, including former Federal Trade Commission Chair Lina Khan, former officials, law professors and advocacy groups. So far she has released an outline, not the text of a bill, and the release lists no co-sponsors. President Donald Trump and House Speaker Mike Johnson have publicly opposed legislative restrictions on the AI industry, according to a news report, and Democrats, her party, are hoping to win control of Congress in the November midterm elections. It is not yet clear how the plan would define an AI company, what a charter would cost, or how today's AI products would fare under its bans.