Former DeepMind researcher warns AI could pose existential risk

A former researcher at Google-owned DeepMind, Bilal Chughtai, said on X and LinkedIn that he believes artificial intelligence could become capable of causing humanity’s extinction if its development continues unchecked. He described the pace of progress since early 2022 as "staggering" and warned that increasingly autonomous AI agents could soon exceed human control.

"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.

Chughtai argued that creating superintelligent systems that surpass human abilities in every domain is plausible within a few years, but current methods for aligning AI behavior with human intentions remain rudimentary. He called alignment "both difficult and unsolved" and said existing training techniques are insufficient to ensure AI acts in humanity’s interest.

Despite the concerns, Chughtai said safe development is possible if technology firms reduce the speed of advancement and address emerging risks before they cause severe harm. He has joined BlueDot Impact, a nonprofit focused on AI safety training.

His remarks follow similar warnings from former Anthropic and OpenAI researcher Jacob Coxon, who also warned that AI developers are "gambling with our lives." Industry leaders, including Anthropic CEO Dario Amodei, have called for a pause on the most powerful models, a stance echoed by OpenAI’s Sam Altman and Elon Musk. Amodei noted that incentives such as potential military advantage make a coordinated slowdown challenging.

President Donald Trump dismissed the warnings as a "hoax" and suggested that only strong political leadership is needed to manage AI risks. The White House, while rejecting broad regulatory measures, is reportedly considering limited safeguards and prefers industry self-regulation to avoid ceding advantage to foreign competitors.