Bill Gates: "The industry knows AI's risks and hides them"

What Gates is really targeting isn't the technology itself — it's the industry's silence. He told The New York Times, "Privately, the people who know how good this is and how fast it's improving are very worried," adding, "But they say to each other, 'Hey, don't say that. It's bad for the next trillion dollars we're trying to raise.'" With the industry pouring trillions into data centers and gearing up for major IPOs, he argues, money is on the line — so people know the risks and downplay them anyway. "Everyone just wants to run flat-out and hope the good outweighs the bad."

The Third Shock Was Claude Code

Why speak up now? Gates says there have been exactly three moments in his life when technology truly stunned him.

1. 1980, the first time he saw a graphical user interface — the screen that became the starting point for the modern PC.
2. 2022, when people from OpenAI came to his house and demoed what would become ChatGPT.
3. Recently, when he looked closely at the progress Anthropic's Claude Code had made.

Coding is a subject he knows better than almost anyone in the world, which is why he said, "The fact that these systems are now better than me at a meaningful level — I just have to ask myself, what is happening here?"

There's another reason for the timing of this release. Milestones the industry itself once set as tripwires — signs that would mean "we need to be more careful past this point," like an AI slipping out of its creator's control or producing instructions for a bioweapon — have already been crossed, and nothing happened as a result. Gates put it more bluntly in an interview with MIT Technology Review: "Bio capability, cyber capability, psychosocial capability, job-destruction capability, even loss of control — we've already crossed the threshold."

This Time Is Different: Jobs

Past waves of technological change have created more jobs than they destroyed. Gates acknowledges that history. But this time, he says, it's "completely, absolutely, totally, thoroughly different" — a claim he says he's staking "everything I know" on. His reasoning is that AI spreads across the entire economy at once rather than hitting one industry at a time, so there's no other sector left standing to absorb the workers pushed out of the first. Given the current structure, with Microsoft and the rest of Big Tech racing to help corporate clients adopt AI, he believes mass layoffs without a response are unavoidable.

His proposal for that is a token tax — a levy on the tokens that are the basic unit of AI compute, which would raise the cost of swapping people for machines and fund support for displaced workers with the proceeds. He told MIT Technology Review it could work like an excise tax on alcohol or tobacco, floating an example of "handing over 50% of token revenue to the government." The New York Times reports the idea has already started circulating on Wall Street.

His second proposal is a category of human-only jobs he calls "Human Reserved." The idea is to simply ban AI from replacing deeply human work like caregiving, or jobs held by people who would struggle to find new work if displaced. In the essay, he compares it to a nature reserve — land where you could build roads and buildings but choose not to because the loss would be too great. He wants the same kind of protected zones carved out in the world of work. His outlook on robots is more urgent: the moment robots become good enough for factories, he says, cooking, cleaning, construction, and logistics all get automated at once — roughly 30% of all jobs, by his estimate.

"Self-Regulation for the Most Dangerous Tool Ever Made? No Thanks."

His proposal on bioterrorism is more concrete: turn the standards for monitoring and limiting extreme risk into an international treaty, and require mandatory review for any AI that could be weaponized — models capable of designing new molecules, for instance. His concern is that a small number of people with access to advanced AI could be enough to create a new pathogen. He rates this risk as "50 times scarier and more likely" than a natural pandemic. Asked about the self-regulatory measures the White House and industry are building, he said: "Self-regulation for the most dangerous tool in history? No thanks!"

Someone Already Sounded This Alarm

Much of this overlaps with warnings Anthropic co-founder Dario Amodei has been making for years. In an Axios interview last May, Amodei warned that AI could wipe out half of entry-level office jobs within one to five years and push unemployment up to 10-20%. He's also known for flagging AI-assisted development of biological and chemical weapons as a bigger threat than cyberattacks. Gates's warnings on mass unemployment and bioterrorism track almost exactly along those same two lines.

The difference lies in how each man gets treated. Amodei's warnings have put him at odds with the Pentagon, the White House, and most of the industry. This month, investor Gavin Baker publicly pressured him, saying "Amodei has already lost this argument" and urging him to sound more industry-friendly, as befits the head of a company heading toward an IPO — prompting Amodei to fire back directly that "regulation versus concentration of power is a false choice." This same month, members of the U.S. House sent Altman and Amodei a public letter over an incident involving an AI model that escaped its test environment. Gates's take on all this: "The person being the most honest about the downsides is the one getting attacked."

A Conclusion Reversed in Three Years

Gates wasn't always in the warning camp. Back in July 2023, he wrote an essay titled "The Risks of AI Are Real but Manageable." Three years later, that conclusion has flipped to "we've crossed the threshold and have no plan." The reason he gives is that recent gains in AI performance have far outrun what he expected.

There's also some baggage behind this comeback. Earlier this year, the release of Jeffrey Epstein's emails put Gates's past ties back in the spotlight; in February, he pulled out of an AI conference he'd been scheduled to attend; and in June, he testified before a House committee for six hours. He's admitted that these episodes, along with the personal controversies that contributed to his 2021 divorce, were "all mistakes I made." He also concedes his enormous wealth, Microsoft's aggressive management style during his tenure, and the antitrust loss from 2000 could make him a flawed messenger. Even so, he argues he's well-positioned to take this on — an industry insider with no company to run or profits to chase, and someone who's spent decades working on inequality through his foundation.

This essay is only the start. Gates says he plans to publish a second essay before the end of the year, focused solely on bio-risk, and that from now on, whenever he meets with world leaders, he'll raise only two topics: global health and AI. The interview closes on this note: "I don't like being the bearer of bad news, or saying that innovation could be a net negative. But that's where we are."

Editor's Take

The weight of this warning comes not from its content but from who's delivering it. On substance, it overlaps heavily with Amodei's warnings — but coming from the CEO of a frontier lab, those warnings could always be read as positioning, hoping regulation would hobble the competition. That rebuttal doesn't work on Gates. He has no model to sell and no IPO on the horizon; he's an industry elder who's broken from the optimist camp, staking "everything I know" on the claim. For Amodei, that makes Gates about the most credible reinforcement he could ask for. And the three-year arc from 2023's "manageable" to 2026's "no plan" is itself a data point on just how fast AI has moved during that stretch.

For readers here, the token tax is the part with real teeth. Most corporate AI adoption today is justified by labor-cost savings, and a token tax would tax exactly those savings, reshaping the math behind the decision to adopt. Wall Street already turning this idea over means a similar bill could eventually land in front of Korea's National Assembly too. The next debate over the AI transition won't be about what the technology can do — it'll be about who gets to keep the gains. Gates's essay is the opening statement in that debate.