Google DeepMind safety researcher Bilal Chughtai has resigned from the company, warning that artificial intelligence could eventually pose an existential threat to humanity. In a post on X, Chughtai said AI had the potential to “kill us all” and suggested that society may be running out of time to prevent a catastrophic outcome.
Chughtai’s resignation comes amid growing concern among AI researchers and industry leaders about the speed at which advanced models are being developed. As AI systems become more capable, experts are increasingly debating whether existing safety measures are sufficient to keep them under human control.
Explaining his decision to leave Google DeepMind, Chughtai said his work had involved ensuring that AI models remained aligned with human interests. His experience inside the company, he said, had made him deeply concerned about the direction in which the technology was moving.
“At Google, I witnessed AI development first hand. I, too, am extremely concerned by the default trajectory of this technology,” he wrote.
Chughtai said the possibility of an AI-driven disaster could no longer be dismissed as a distant concern. He warned that humanity might not have enough time to develop safeguards before increasingly powerful systems are deployed.
“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,” he added.
Another DeepMind Researcher Resigns
Chughtai is not the only Google DeepMind researcher to leave over concerns about advanced AI. Josh Engels, who worked on the company’s artificial general intelligence safety team, has also resigned.
Engels said he believed there was a “terrifying chance” that AI systems could cause immense harm within the next five years. Although he acknowledged uncertainty about the exact probability, he said the risk was serious enough to make AI safety the most important problem in the world.
Engels has since joined Model Evaluation and Threat Research, or METR, a nonprofit organisation that evaluates the capabilities and risks of frontier AI models. He said more independent organisations were needed to hold AI companies accountable and study the dangers of advanced systems from different perspectives.
The departures come as AI safety becomes a central issue across the technology industry.
AI’s Rapid Progress Raises Alarm
Chughtai described the pace of AI development as extraordinary. He said that when he began working in the field in early 2022, AI systems were “amusingly useless”. Just four years later, he argued, AI agents were solving difficult mathematical problems and showing signs of behaviour that raised serious safety questions.
He referred to claims that OpenAI systems had solved the Navier-Stokes Millennium Challenge, a long-standing mathematical problem. Mathematicians, however, questioned whether the system may have relied on existing research data.
Chughtai also pointed to reports that hundreds of AI agents had attempted to hack the third-party platform Hugging Face while trying to manipulate an evaluation. Such incidents, he suggested, demonstrated how difficult it could become to control increasingly autonomous systems.
According to Chughtai, future AI systems could become “superintelligent”, exceeding human capabilities across multiple fields. His biggest concern was that these systems might take dangerous actions beyond human control, potentially resulting in the permanent loss of human power or even humanity’s destruction.
Calls for a Slower AI Race
Chughtai argued that AI companies must urgently improve alignment research, which focuses on ensuring that AI systems behave in accordance with human goals and values.
He said current understanding of how to train AI systems to reliably pursue human interests remained extremely limited. At the same time, he warned that AI capabilities were improving much faster than safety research.
He also backed calls for greater coordination among AI companies and governments to slow down development.
“We need to coordinate to avoid this manic race between AI companies,” Chughtai said, arguing that AI progress should move at a pace society can manage.
The debate has also reached the political sphere. While AI leaders such as Anthropic CEO Dario Amodei have urged companies to slow development, US President Donald Trump has opposed such proposals. Trump has argued that the United States must continue advancing AI to compete with China.
“Whoever wins AI wins,” Trump said, reflecting the growing tension between technological competition and concerns over AI safety.
