A former Google DeepMind employee has become the latest voice in tech to say plainly that AI could kill all humans, and that the window to prevent it may already be closing. Bilal Chughtai, who left the company in July after working on AGI safety and alignment, posted the warning on X on Monday. It landed just days after a separate resignation from an Anthropic researcher raised similar alarms about where the technology is headed.
Coming so soon after that first warning, Chughtai’s comments have pushed AI existential risk back into the center of public conversation. Two people who worked inside two of the world’s leading AI labs are now saying, on the record, that the technology they helped build could end up killing everyone.
Background: A Pattern of Warnings From Inside the Industry
Concern about artificial intelligence danger is not new, but the tone has shifted. For years, warnings about AI risk came mostly from outside commentators, philosophers, and a handful of academics. Lately, the loudest voices have been people who actually built the systems.
The immediate spark was Jacob Coxon, a researcher who had worked at both Anthropic and OpenAI. Last week, Coxon announced his resignation and said the people building AI privately believe it could kill everyone by the end of the decade. He described the current approach as a race that companies are running without anyone in control of the finish line.
Anthropic scientist Evan Hubinger responded soon after, backing Coxon’s account and putting a number on it: he estimated better than a 10 percent chance that AI causes human extinction within ten years. That figure, coming from someone still working at a major AI safety lab, is part of what turned this into a bigger story.
Details: What Bilal Chughtai Actually Said
Chughtai worked as a research engineer at Google DeepMind and co-authored several research papers before leaving the company. On his personal website and on X, he laid out his reasoning for going public.
He said he sincerely believes AI has the potential to kill everyone, and that time may be running out to change course. He was not vague about the stakes, and he did not hedge the claim the way many industry statements tend to.
Chughtai has since moved to BlueDot Impact, a nonprofit focused on AI safety education, where he continues working on the same alignment questions he studied at DeepMind. He said avoiding catastrophe is still possible, but only if companies stop treating safety as an afterthought in a race for market share.
He was also specific about what needs to change. He called for pacing AI development to a speed society can actually absorb, so that new risks can be caught and addressed before they cause serious harm. That is a different ask than simply stopping AI research; it is a call for slower, more deliberate progress.
Quotes: What the Researchers Are Saying
Chughtai summed up his position in blunt terms, writing that AI could genuinely “kill us all” and that there may be little time left to prevent it. It’s a short line, but it carries real weight coming from someone who spent years working on the safety side of frontier AI.
He was equally direct about the industry’s competitive dynamics, saying safe development requires coordination “to avoid this manic race between AI companies.” That phrase points at a specific worry: that safety gets sacrificed whenever one lab feels pressure to ship faster than a rival.
Anthropic CEO Dario Amodei added his own voice over the weekend, calling for the pace of advanced AI development to slow down. Notably, he found some unlikely allies. Elon Musk and OpenAI’s Sam Altman both offered rare public backing for that position, even though their companies are usually described as competitors in the same race Chughtai warned about.
Not everyone in a position of power agrees. President Donald Trump has dismissed the warnings outright, calling the AI industry’s push for regulation a “hoax” on social media. That single line captures how divided the response to these AI safety warnings has become, even among people who are paying close attention.
Impact: Why This Warning Is Landing Differently
Individually, none of these statements would be shocking. Researchers have raised alarms about AI risk before, and the field has weathered plenty of doom-laden headlines. What makes this moment different is the clustering: two resignations, two public statements, and a specific probability estimate, all within about a week.
For the AI industry, this creates a genuine problem. Companies like Google, Anthropic, and OpenAI depend on public trust and government goodwill to keep operating with minimal restriction. When their own former and current researchers say the technology could kill everyone, that trust gets harder to maintain, regardless of how leadership responds.
For governments, the pressure to act is building, but the response so far is fractured. Some officials are pushing for tighter oversight of frontier AI systems, while others, including the U.S. president, are waving the concern away as exaggerated or politically motivated. That split makes coordinated global regulation harder to imagine in the near term.
For the public, the practical effect is more uncertainty than clarity. Hearing insiders debate whether AI could kill all humans is unsettling on its own, and the lack of agreement among experts makes it difficult for ordinary people to gauge how seriously to take the warning.
Conclusion: What Comes Next
Chughtai’s resignation and warning are unlikely to be the last of their kind. If Coxon and Chughtai both left major labs within weeks of each other citing the same core concern, it suggests there may be more researchers weighing similar decisions behind closed doors.
The bigger question is whether any of this changes how AI companies actually operate. Calls to slow down, including from Amodei, Musk, and Altman, have not yet translated into a public commitment or timeline from any major lab. Whether that changes will likely depend on how the public and regulators respond in the coming weeks.
For now, the debate over AI existential risk remains split between researchers warning of catastrophe, executives calling for caution while continuing to build, and political leaders dismissing the concern altogether. That gap between warning and action is where this story will keep playing out.
Frequently Asked Questions
What did Bill Gates warn about AI?
Bill Gates has shifted from cautious optimism to open concern about artificial intelligence risks. In a lengthy essay on his personal website, he warned that AI is advancing faster than governments can prepare for, and that no country is currently ready for the disruption it will cause. Gates pointed to three major dangers: widespread job losses across law, medicine, software, and manufacturing; AI being used to scale up cyberattacks, bioterrorism, and fraud; and long-term damage to human relationships and child development from over-reliance on AI systems. In interviews since, he has said governments are “way behind” on managing these risks and has been raising the issue directly with world leaders, including U.S. President Donald Trump.
Can AI harm a human?
Yes, in ways that are already documented rather than purely hypothetical. Current AI systems have been used to generate convincing scams, deepfakes, and disinformation that cause real financial and reputational harm. Security researchers have also flagged AI’s growing role in lowering the skill barrier for cyberattacks and, according to Bill Gates and others, potentially assisting with the creation of bioweapons. The warnings from researchers like Bilal Chughtai and Evan Hubinger go further, arguing that future, more capable AI systems could pose a direct existential threat to humanity as a whole if they are deployed without adequate safeguards or if they act in ways their creators did not intend and cannot control.
What did Elon Musk say about AI?
Elon Musk has repeatedly described artificial intelligence as one of the most significant risks facing humanity, at times saying it poses a bigger long-term danger than nuclear weapons. He has called for regulatory oversight of advanced AI development for years, arguing that the technology needs guardrails before it becomes too powerful to safely rein in. Musk’s recent public support for Dario Amodei’s call to slow the pace of advanced AI development is notable because it puts him on the same side as Anthropic, a company he does not run, on an issue where AI labs are usually seen as competing rather than cooperating.
