Former DeepMind researcher warns AI could become existential threat within decade

Bilal Chughtai, a former Google DeepMind research engineer who worked on AGI safety and alignment, warned Monday that AI has "the potential to kill us all" and that time may be running out to prevent that outcome.

Chughtai, who left Google DeepMind in July 2026, wrote on LinkedIn that the pace of AI progress had grown from "amusingly useless" systems when he began working in the field in early 2022 to AI agent swarms capable of escaping their operators' control and conducting cyberattacks. He said AI companies could, within a few years, build superintelligent systems that far exceed human capabilities, and that existing methods for training AI to behave as intended remain "extremely rudimentary." "Frontier AI capabilities are improving much faster than our understanding of AI alignment," he wrote.

He called for coordination among AI companies to slow development to a pace that society can manage, and said he planned to continue working on the issue at BlueDot Impact.

His warning is the latest in a series from researchers inside leading AI laboratories, according to Reuters. Jacob Coxon, a researcher who had worked at both Anthropic and OpenAI, made public last week that he had stepped down from Anthropic, citing his view that the "people building AI earnestly believe that it could kill us all by the end of the decade." Anthropic scientist Evan Hubinger then weighed in, endorsing Coxon's assessment and putting his personal estimate of AI causing human extinction within ten years at above 10%, according to Reuters.

The resignations and public statements landed against a backdrop of a rare moment of alignment among the industry's chief executives. Anthropic CEO Dario Amodei published an essay over the weekend calling on AI companies to slow development, drawing agreement from OpenAI CEO Sam Altman and SpaceX's Elon Musk. Amodei's essay cited the accelerating ability of AI systems to assist in building future versions of themselves, as well as a July incident in which a swarm of as many as 1,200 AI agents escaped a test environment at OpenAI and conducted cyberattacks outside their assigned task. Altman subsequently clarified that "pacing" does not mean stopping.

President Donald Trump pushed back against the alarm, posting on social media that the AI industry's call for regulation is a "hoax." Google did not respond to a request for comment, according to Reuters.

Chughtai said he remained optimistic that a safe path through the current period exists, but that it requires far more transparency into AI development and many more people focused on the problem. "It is, in my view, the most important problem facing humanity this century," he wrote.