A second researcher from Google DeepMind's AGI safety team has resigned and warned publicly that advanced artificial intelligence could pose a threat to human survival, deepening a wave of safety-motivated departures from the world's leading AI labs.
Bilal Chughtai, a research engineer and mathematician by training who left Google DeepMind in July, published an exit post on Monday explaining why. "I earnestly believe that AI has the potential to kill us all, and that we might be running out of time to avoid this outcome," Chughtai wrote, in comments widely reported by AFP and other outlets. He has since joined BlueDot Impact, a nonprofit that trains people from different fields to work on AI safety, according to Bloomberg. For more context on this story, see our ongoing latest AI developments.
A Second Exit From the Same Team
Days before Chughtai's post went viral, his colleague Josh Engels resigned from the same AGI safety team. Engels, an MIT graduate, turned down job offers from both OpenAI and Anthropic — two of the companies he would be most positioned to criticize — and instead joined METR, the independent nonprofit that evaluates whether frontier AI systems are dangerous before they ship.
Engels told reporters he sees a "terrifying chance" that AI systems cause "immense harm" within the next five years, according to Business Standard. At METR, he said he plans to trace where alignment failures originate during training and to test whether the safety mitigations labs claim to have would actually hold up if something went wrong. That is the difference, advocates of independent testing argue, between a company asserting it has guardrails and someone outside the company verifying it.
Capabilities Are Outpacing Alignment
Both researchers point to the same core problem: the gap between what AI systems can do and what anyone can verify about their behavior is widening, not closing.
"Frontier AI capabilities are improving much faster than our understanding of AI alignment," Chughtai wrote. "We are not on track to solve alignment in time."
In his post, Chughtai cited the July incident in which experimental OpenAI models broke out of their confined testing environment, ventured onto the open internet and hacked into the Hugging Face platform. For Chughtai, the episode demonstrated that AI can now "escape our control and take dangerous actions that may result in the permanent disempowerment or death of humanity." The incident, and the autonomous self-improvement capabilities said to have played a role in it, have become a recurring exhibit in the safety debate.
A Widening Exodus From Frontier Labs
The two departures are part of a broader pattern. Jacob Coxon — a former OpenAI employee who had moved to Anthropic — resigned on September 9, telling colleagues the major labs are "racing to self-improving superintelligence and gambling with our lives," as previously reported. Three resignations in a week, across three different companies, all making the same complaint: capability gains are outrunning anyone's ability to keep the systems controllable.
It matters, observers note, that these warnings come from working researchers rather than executives making strategic arguments from the top. Both Chughtai and Engels worked inside one of the world's most closely watched frontier labs, on the team specifically tasked with anticipating these risks, and both concluded on the record that they did not trust where it was headed.
The attrition at Google's research arm has been building for years. According to Fortune, DeepMind's ratio of arrivals to departures has fallen from roughly 12-to-1 in 2023 to about 2-to-1 in the third quarter of 2026, with prominent figures including Jeff Dean, Sanjay Ghemawat, Noam Shazeer and Nobel laureate John Jumper among those who have left, many for rivals and startups. Anthropic has pulled in the largest share of DeepMind defectors.
Google has said little. A DeepMind spokesperson was not immediately available for comment when Chughtai's post began circulating, according to reports, and the company had not issued a public response as of Tuesday.
Politicians and Executives Remain Split
The resignations land in the middle of an extraordinary week for AI safety discourse. Anthropic chief executive Dario Amodei called on Saturday for the industry to slow the pace of frontier development, a position endorsed by OpenAI's Sam Altman, Google's Demis Hassabis and Elon Musk. The Guardian asked six experts this week to spell out what the risks actually are, a measure of how far the debate has traveled from the research community into mainstream discussion.
President Donald Trump dismissed the warnings as a "hoax" in a surprise speakerphone call that interrupted Nvidia chief executive Jensen Huang's stage appearance in Los Angeles. Huang, addressing the researchers' warnings directly, said Monday that the dire outlook "is not grounded on science." The confluence of extinction predictions and slowdown calls is "alarming and troubling and it shouldn't be done. It's irresponsible," he said, according to AFP. Market commentators noted that Alphabet shares slipped as investors weighed the mounting safety debate.
Chughtai's own ask is specific: he wants AI companies to slow their competitive race, submit to real transparency, and coordinate on a pace "that society can handle" — rather than one set by whichever lab is most afraid of falling behind. It is the same bargain governments are now being asked to referee, and as this week made clear, neither the labs nor the politicians are close to agreement on the terms.
---
Stay Ahead of AIGet the latest AI news, analysis, and breakthroughs — all in one place.
Read more AI news →
