Bill Gates tells Meet the Press: regulate AI — it’s powerful enough for a billion deaths

25 Sep 2026 (ET): Microsoft co-founder Bill Gates told NBC’s Meet the Press that AI companies cannot adequately self-regulate, that Congress should pass legislation, and that AI is “certainly powerful enough to drive events that… cause a billion deaths” when paired with bad actors.

POLITICS desk — Bill Gates co-founded Microsoft, and he told a national news show that AI companies cannot police themselves. Asked if Congress should pass a law, he said absolutely, and he said the technology is already powerful enough, paired with people who want to do harm, to drive events that kill a billion people. That answer matters while the Speaker of the House has said Congress should not rush a federal bill, and while some of the people building the systems still want each lab to set its own pace. These are remarks NBC reported. They are not a statute.

What the lede says he asked for, and which day the page calls Wednesday. Microsoft co-founder and billionaire philanthropist Bill Gates on Wednesday called for lawmakers in Washington to regulate the development of artificial intelligence, weeks after the CEOs of some of the nation’s biggest AI labs agreed on the need to slow the pace of AI progress. The article is dated Friday, 25 Sep 2026. Wednesday, read against that Friday, is 23 Sep 2026. That reading is this desk’s. The page does not print 23 Sep. Regulate, here, means pass rules the companies have to follow, not a suggestion they can ignore. That gloss is this desk’s. Weeks after, and some of the nation’s biggest AI labs, are the lede’s. The names come in the next paragraph. This desk did not add a lab the lede does not name.

The sentence on self-regulation, and the one-word answer on a law. “No one thinks self-regulation is enough,” Gates told NBC News’ Meet the Press in an interview set to air in full Sunday. Self-regulation means the companies write and police their own rules. That gloss is this desk’s. Sunday, read against a Friday 25 Sep publication, is 27 Sep 2026. That reading is this desk’s. Set to air is future tense on the page. The full interview had not aired when NBC published at 7:00 a.m. Eastern on Friday, and it had not aired when this desk filed on Friday night Eastern. Asked by moderator Kristen Welker whether there needs to be legislation passed in Washington, Gates said, “Absolutely.” Legislation is a law Congress passes. That gloss is this desk’s. Absolutely is his answer to Welker’s question, as NBC prints it. It is not a bill number.

What he said the safeguards have to be. “You need law enforcement and the politicians to get into the discussion about what safeguards and monitoring look like,” he said. “And that has to be a required thing. And it will be a little bit of overhead for the industry, but not a dramatic slowing of what they’re doing.” Safeguards are limits on what the systems are allowed to do. Monitoring means someone checks whether those limits hold. Overhead, in his sentence, is extra cost and extra process for the companies. Those glosses are this desk’s. Required, a little bit of overhead, and not a dramatic slowing are his words, as NBC quotes them. They are not a statute this desk read.

The scale he put on misuse, kept separate from a prediction of certain extinction. In Wednesday’s interview, NBC writes, Gates didn’t downplay the risk that AI could wipe out humanity in several years, but called for Americans not to ignore the risks that AI could be used by bad actors to cause large-scale harm as soon as today. Didn’t downplay, and as soon as today, are NBC’s account of his emphasis. They are not a sentence in quotation marks. A bad actor, here, is a person who means to cause harm. That gloss is this desk’s. Then the quotation: “AI is certainly powerful enough to drive events that, you know, cause a billion deaths. You know, so even though it’s pretty hard to get to 100%, there’s never been a weapon as powerful as the combination of people with ill intent using the latest AI tools,” he said. A billion deaths is his scale for events the tools could drive. Pretty hard to get to 100% is also his, in the same quotation. One hundred percent, in that sentence, is the far end he says is hard to reach. It is not a forecast that every person dies. Ill intent means the person using the tool wants to do harm. That gloss is this desk’s. NBC’s subhead says harm billions of people. The quotation says a billion deaths. Do not file either line as a prediction that humanity is certain to go extinct.

Who NBC says had already asked to slow down, and what the whistleblower said. Earlier this month, the CEOs of Anthropic and OpenAI, two of the nation’s biggest AI labs, agreed that the race to develop powerful AI systems should slow down. Earlier this month is the page’s clock. It does not print the day. Their statements came after a whistleblower who recently resigned from Anthropic, Jacob Coxon, publicly declared that “neither company is acting responsibly. They are racing straight to self-improving superintelligence and gambling with our lives.” A whistleblower, here, is an insider who speaks out about the employer. Superintelligence, in his sentence, is a system that improves itself past what people can control. Those glosses are this desk’s. “The people building AI earnestly believe that it could kill us all by the end of the decade,” Coxon added in a series of social media posts and interviews that week. Kill us all, and the end of the decade, are Coxon’s, as NBC prints them. They are not Gates’ quotation. Do not move them onto Gates.

What Coxon told the same program, which is a different sentence from the posts. In an interview with Meet the Press earlier this month, Coxon said that AI companies should be allowed to regulate themselves. “I do think that the people running the labs, especially their recent communications, are completely genuine. They would like to slow themselves down,” he said. Allowed to regulate themselves is NBC’s account of his position. Completely genuine, and would like to slow themselves down, are his. Earlier this month is the page’s clock for that interview too. It does not print the day. That position is Coxon’s. Gates’ sentence on the same program is that no one thinks self-regulation is enough. Do not collapse them.

What the Speaker said, several weeks earlier on the same program. Appearing earlier on Meet the Press several weeks ago, Speaker Mike Johnson, R-La., declined to bring House lawmakers back to Washington to pass legislation to regulate the pace of AI growth, saying that he would defer to the AI companies. Several weeks ago is the page’s clock. It does not print the day. Declined, and defer to the AI companies, are NBC’s account. “We have to put some guardrails, some safety measures in place to ensure that AI doesn’t run away,” the speaker said, adding that while AI CEOs had called for a slowdown, they hadn’t yet offered a path forward that involved federal legislation. Guardrails, in that sentence, means limits so the technology does not get away from the people running it. That gloss is this desk’s. “They don’t themselves even have the answer as of today. So we can’t rush in and have Congress do something that would be detrimental to national security,” Johnson said. Detrimental means harmful. National security means the country’s defense. Those glosses are this desk’s. As of today, in his sentence, is the day he spoke, not 25 Sep. This desk did not hear the interview.

What NBC quotes from Dario Amodei’s essay, which is not a Gates sentence. In his essay calling for a slowdown in the pace of AI development, Anthropic CEO Dario Amodei specifically called for global and federal cooperation. “The most effective method of pacing is via regulation that targets all U.S. frontier AI companies, as that covers even those who are unwilling to cooperate voluntarily,” he wrote, adding later in the piece: “Global pacing will require cooperation with China, the autocratic country with by far the most advanced AI capabilities.” Pacing, here, means slowing the race on purpose. Frontier AI companies, in his sentence, means the labs building the most advanced systems. Autocratic means a government that is not a democracy. Those glosses are this desk’s. Those two sentences are Amodei’s, as NBC quotes the essay. This desk did not read the essay on this filing. Do not file them as Gates’ words.

Altman at the United Nations, and Zuckerberg’s disagreement, both as NBC places them. Sam Altman, the CEO of OpenAI, on Wednesday spoke before a gathering of the United Nations, calling for international standards for “measuring capabilities, assessing risks, determining whether safeguards are sufficient and preserving meaningful human oversight” over AI. Wednesday is the same word the page uses for Gates’ interview. This desk does not print a second date the page does not print. Standards, here, means shared rules for how to measure the systems and how to tell if the safety checks are enough. Human oversight means a person still has to be able to watch and stop the system. Those glosses are this desk’s. The words inside the quotation marks are Altman’s call, as NBC prints it. Meta founder and CEO Mark Zuckerberg disagreed with that approach. “I don’t think that we need some kind of industrywide coordination,” he told NBC News in an interview published Thursday. “I think that each lab needs to take the time, and when it sees that there are issues, you just take the time that you need internally to basically make sure that you’re proceeding safely.” Industrywide coordination means the labs moving together, not each lab deciding alone. That gloss is this desk’s. Thursday, read against Friday 25 Sep, is 24 Sep 2026. That reading is this desk’s. The page says published Thursday. It does not say he spoke on Thursday.

What NBC says the states have started, without turning “last week” or “this week” into a calendar day the page does not print. At the state level, some governors have already begun the process of regulating AI companies and the data centers they use to power their development. A data center is a building full of the computers that train and run AI. That gloss is this desk’s. In California, Gov. Gavin Newsom last week signed an executive order creating a board of experts tasked with creating a plan for the state to strengthen its laws on AI safety. An executive order is a governor’s directive. It is not a statute Congress passed. That gloss is this desk’s. Last week is the page’s clock. This week, in Maryland and New York, Govs. Wes Moore and Kathy Hochul each also released plans aimed at boosting AI regulation in their states. Moore’s order creates a “three pronged approach” to regulation, while Hochul’s creates a new Office of Digital Innovation, Governance, Integrity and Trust to oversee legal compliance for AI companies. Three pronged is NBC’s spelling, without a hyphen. Legal compliance means following the law that applies to the company. That gloss is this desk’s. The page does not print the text of the orders. This desk did not read them.

What Microsoft’s president said the same week, which is not Gates’ interview. Speaking at a United Nations session on AI and human development this week, Microsoft President Brad Smith rejected the notion that AI companies have to wait for federal regulation to slow themselves down and criticized them for constantly racing against each other to make more advanced AI products. This week is the page’s clock. “Imagine for a moment that there’s only two big companies in the world that produce commercial airplanes,” Smith said. “Imagine that somebody comes into your office and says, ‘I know we’re in a race to get out the next airplane first, but we’re worried that the airplane will crash and people will die.’ Would you say, ‘I need to win this race’?” Smith asked. “Or, would you say, ‘I’m going to put the safety of my passengers and the safety of the public ahead of my winning that race’? That is the fundamental question.” Those sentences are Smith’s, as NBC quotes them. Gates co-founded Microsoft. Smith is the company’s president. Do not collapse the two men into one interview.

What the card shows. The card is the NBC photograph of Bill Gates from this Meet the Press interview, the still on the article. He has short gray hair and black-rimmed glasses. He wears a navy sweater over a white collared shirt, with a small microphone at the collar. He looks slightly to the right of the camera. Behind him is a bright window. At the lower right, a bouquet of orange and green flowers is out of focus. No date is printed on the card. It is not a stock portrait and not a wordmark. The catalog chip is POLITICS.

Plain English for the rest of the card. Bill Gates, who co-founded Microsoft, told NBC’s Meet the Press that AI companies cannot adequately police themselves. Asked if Washington needs to pass a law, he said absolutely. He said law enforcement and politicians have to help decide what the limits and the watching look like, that those limits have to be required, and that the extra process will be a little overhead, not a dramatic slowdown. He said AI is powerful enough, in the hands of people who want to do harm, to drive events that cause a billion deaths, and that reaching 100 percent is pretty hard. He did not say extinction is certain. NBC’s story also records the labs’ slowdown calls, a former Anthropic employee’s warnings, the Speaker declining to rush a House bill, state plans in California, Maryland, and New York, Sam Altman asking the United Nations for standards, Mark Zuckerberg saying each lab can decide alone, and Microsoft’s president comparing the race to an airplane nobody should ship if it might crash. The full interview was set to air Sunday. The card is Gates in that interview.

REPORTED here: NBC News’ 25 Sep 2026 story by Alexandra Marquez, stamp Sept. 25, 2026, 7:00 AM EDT, of Bill Gates on Meet the Press — Tier B news report of an interview, not a statute. STATUS REPORTED. The desk label on the chip is REPORTED. The Wednesday call for Washington to regulate, the self-regulation quotation, the Absolutely answer to Kristen Welker, the safeguards and monitoring quotation, NBC’s didn’t-downplay sentence, the billion-deaths quotation including pretty hard to get to 100%, the subhead’s harm billions of people, the Anthropic and OpenAI slowdown, both Coxon quotations and his self-regulation position, Speaker Mike Johnson’s decline and both of his quotations, both Amodei essay quotations, Altman’s United Nations standards call, both Zuckerberg quotations, Newsom’s executive order, Moore’s three pronged approach, Hochul’s new office, and Brad Smith’s airplane quotations are NBC’s. The Verge and Newsweek pages this desk also read carry the same Gates quotations and do not add a law. Forbes returned HTTP 403, so this filing does not quote a Forbes body. NOT claimed: that Gates predicted certain human extinction, that 100% in his quotation is a forecast that everyone dies, that billions of people harmed and a billion deaths are the same count, that a bill was introduced or passed, that this desk watched the Sunday broadcast or a video short, that Wednesday’s calendar date is printed on the page, a Forbes sentence, Newsweek’s White House summit, poll, Pope, or other leaders as if they were in the NBC story, a stock tip, or investment advice. The card is the NBC interview still. Distinct from the already-filed ftc-ai-agent-liability, pentagon-anthropic-supply-chain, and fed-schmid-ai-too-big-to-fail. The catalog chip is POLITICS, not SIGNAL.