There’s a New Warning From Bill Gates That Could Be His Most Serious Yet

One billion. That is the figure Bill Gates put on the table Sunday during his sit-down with NBC’s Kristen Welker on Meet the Press. Pressed on whether artificial intelligence could ultimately end all of humanity, the Microsoft co-founder sidestepped that exact question and pointed to a ten-digit death toll. It is a number usually tied to pandemics and nuclear war, and Gates attached it to software. His explanation of how came next.

Gates delivered the line himself: “AI is certainly powerful enough to drive events that, you know, cause a billion deaths.” He acknowledged that a full 100 percent is a steep bar, then set the technology against every weapon in history: “there’s never been a weapon as powerful as the combination of people with ill intent using the latest AI tools.”

Gates named a remedy as directly as he named the risk. Law enforcement and elected officials, he said, need a seat in the conversation about safeguards. He also called for keeping watch on every advanced model and maintaining an exact log of its activity. Someone with legal authority would have to hold that record, which is a striking request from a man who spent decades near the center of the industry he now wants supervised.

Anthropic’s September Report Lists Drone Swarm Code and Five Blocked Bioweapons-Related Attempts

The kind of misuse Gates described already has a paper trail. On Sept. 10, Anthropic released its latest threat intelligence report, covering December 2025 through August 2026. Analysts sorted the abuse into seven categories: cyber operations, influence campaigns, surveillance, fraud, biological misuse, conventional weapons, and illicit copying of models. The company says it disrupted every operation described. The weapons cases drew the most scrutiny.

One case involved freelance actors believed to be based in Russia. According to Anthropic, they used Claude Code to test and refine software for a swarm of first-person-view kamikaze drones. The code included what the report calls autonomous lethal engagement, meaning onboard systems could spot a target and order detonation with no human in the loop. The team trained its recognition system on scraped footage from the war in Ukraine, sorting targets into enemy and friendly.

The biology cases came from a different kind of user. Anthropic identified and blocked five attempts by scientists to use Claude for research that could support biological weapons work. One came from a researcher tied to a military institute who wanted help with a gain-of-function study. Another user tried to draft a state-sponsored grant proposal aimed at making the chikungunya virus spread more easily and slip past immune defenses. Anthropic responded by restricting a wide range of dual-use biology queries on Claude Fable 5 and later models.

Amodei’s “Pace the Frontier” Essay Drew Public Backing From Sam Altman and Elon Musk

The unease had been building inside the labs before any chief executive spoke up. Jacob Coxon, a former Anthropic researcher, resigned and published a post that spread widely online, cautioning that the companies racing to build ever stronger AI were putting the public at risk. Coverage describes him as warning of an existential threat. His departure came within days of an essay from his former boss.

Anthropic CEO Dario Amodei published “We Must Pace the Frontier” on Sept. 12, arguing that AI companies should deliberately slow how quickly their models gain new abilities. He pointed to two triggers: AI increasingly building the next generation of AI, and the OpenAI-Hugging Face incident, in which a swarm of agents attacked targets it was never assigned. Pacing should not halt training, he wrote, and regulation covering every US frontier lab is the strongest way to enforce it.

Reaction came fast. OpenAI’s Sam Altman publicly agreed, and Elon Musk wrote on X that “Dario is right.” Gates had reached a similar place weeks earlier. In an August essay, he wrote that no plan exists for easing the world into the AI era, and that he would probably back any credible plan for slowing the technology globally. That was a sharp turn for a man long known as an optimist. Washington’s answer came in a Truth Social post.

Trump Mocked Amodei as a “Perfect Little Angel” as Nadella Kept Enforcement in the Labs’ Hands

President Donald Trump answered Amodei on Sept. 14, mocking the Anthropic chief as someone posing as a “perfect little angel.” The president has shown little appetite for regulating AI, though he announced an “AI force” that he said would “watch over” the industry. The two men were due to share dinner at the White House on Sunday night, the same day Gates went on air.

Gates advises Microsoft CEO Satya Nadella, whose public position leaves more room for the labs to police themselves. Nadella told the All-In Podcast that when engineers spot a show-stopper, meaning flaws in code that could produce rogue agents, the answer is to “stop the show.” In a September memo to employees reported by Business Insider, he said companies must give safety the time it demands. Nearly everyone at the top of the industry accepts that stronger guardrails are needed. The fight is over who holds the enforcement power.

When Trump waved off catastrophic AI fears as a hoax, Gates took issue with it directly: “It’s not a hoax at all.” He also rejected the idea that rules would cost America its footing against China, saying government safeguards would not hamper the country in that contest. His remedy runs through Congress, where he wants safety legislation passed. Gates advises the CEO of one of the world’s largest AI builders, and the rules he is asking for would bind that company too.