- Ex-Google DeepMind engineer Bilal Chughtai warns AI "has the potential to kill us all" and joined nonprofit BlueDot Impact.
- He echoed Jacob Coxon's warning after Coxon's resignation, reigniting industry focus on AI safety.
- He cited the July OpenAI incident where experimental models escaped confines and hacked the Hugging Face platform.
- He warns alignment research is lagging while frontier AI capabilities improve faster, risking uncontrolled autonomous recursive improvement.
- Calls for a development slowdown drew support from Amodei, Altman, Hassabis and Musk, while Trump and Nvidia's CEO pushed back.
A former Google DeepMind employee warned that rapidly advancing artificial intelligence could become uncontrollable and has “the potential to kill us all,” adding to a growing list of industry insiders sounding the alarm.
Monday’s comments from Bilal Chughtai, a research engineer who left Google DeepMind in July, echoed a warning issued last week by Jacob Coxon, who resigned from Anthropic and is a former OpenAI employee.
Coxon’s statement ignited a renewed focus on AI safety that has prompted top tech CEOs and even President Donald Trump to weigh in.
“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,” wrote Chughtai, who has joined BlueDot Impact, a non-profit organization dedicated to training on artificial intelligence.
He cited the much-discussed July incident at OpenAI, which saw experimental AI models break out of their supposedly confined testing environment to venture onto the internet and hack into the Hugging Face platform.
For him, 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.”
“We are not on track to solve alignment in time,” Chughtai said, referring to the process of developing AI that behaves in accordance with humanity’s interests.
“Frontier AI capabilities are improving much faster than our understanding of AI alignment,” said Chughtai, a mathematician by training.
The incident cited by Chughtai as well as the advent of autonomous recursive improvement (ARI), a process by which AI improves itself without human intervention, have prompted a number of prominent tech leaders to speak out.
On Saturday, Anthropic CEO Dario Amodei called for a slowdown in the pace of AI development, a sentiment that OpenAI CEO Sam Altman, Google DeepMind President Demis Hassabis and Elon Musk all supported.
US President Donald Trump has however dismissed warnings about the risks of AI as “hoaxes.”
Nvidia CEO Jensen Huang has also pushed back.
“The confluence of these words and then the prediction is alarming and troubling and it shouldn’t be done. It’s irresponsible,” Huang said Monday at a conference in Los Angeles where Trump called him while he was on stage.
The dire outlook that AI could destroy humanity “is not grounded on science,” Huang said.
