A former researcher at Google DeepMind has issued a stark warning about the potential dangers of artificial intelligence, stating that the technology could ultimately lead to humanity's demise. Bilal Chughtai, who recently left the prominent AI lab after working on AGI safety and alignment, expressed grave concerns about the rapid pace of AI development, which he has observed firsthand.
"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 on X and LinkedIn. He noted that the progress since he entered the field in early 2022 has been "staggering," particularly the development of increasingly autonomous AI agents that developers may soon be unable to reliably control.
Chughtai elaborated that AI companies could succeed in creating superintelligent systems that surpass human capabilities in all domains within the next few years. He expressed uncertainty about whether these systems would align with human intentions, warning that misaligned AI could escape control and lead to the "permanent disempowerment or death of humanity." He described alignment—the effort to prevent such outcomes—as both a difficult and unsolved problem, with current methods for ensuring AI systems act in accordance with human desires being "extremely rudimentary."
Despite these concerns, Chughtai believes that AI can still be developed safely if technology firms slow down their "manic race" to create more powerful systems. He advocates for pacing AI development to a speed that society can manage, allowing emerging risks to be addressed before significant harm occurs. Chughtai has since joined BlueDot Impact, a nonprofit focused on AI safety training.
His warning follows similar sentiments from other AI researchers. Jacob Coxon, a former researcher at Anthropic and OpenAI, recently resigned, accusing companies of "gambling with our lives." Coxon also noted that many developers believe AI could pose an existential threat by the end of the decade.
These concerns have also reached the executive level. Anthropic CEO Dario Amodei has called for the industry to slow down the development of its most advanced AI models, a sentiment echoed by OpenAI's Sam Altman and Elon Musk. Amodei acknowledged the difficulty in coordinating such a slowdown due to the significant military and competitive advantages offered by rapid AI advancement, stating, "I don’t know if it’s possible, but we should try."
Former President Donald Trump, however, has dismissed these warnings as a "HOAX" and a "SICK conspiracy." He argued that AI fears are being used against AI and data centers to benefit China. Trump emphasized the strategic importance of the U.S. maintaining its lead in AI, asserting that the only necessary controls are a "STRONG AND SMART (High IQ!) PRESIDENT."
Trump criticized Amodei, suggesting the Anthropic CEO was feigning innocence while asserting that existing government powers are sufficient to regulate AI companies. Meanwhile, the White House has reportedly been considering "limited safeguards" for AI, though not a broad regulatory crackdown. One executive familiar with the administration's thinking indicated that any agreed-upon safeguards would be minimal, reflecting concerns that wider restrictions could cede an advantage to China. The administration appears to favor industry self-regulation and narrowly focused guardrails. Google has been contacted for comment.