Advanced AI systems could pose existential risks to humanity, and losing control over them may lead to catastrophic outcomes. This warning comes from Nobel Prize-winning physicist Jeffrey Hinton, a pioneer in the field, as reported by The Times.
"It would be exceedingly unwise to develop superintelligence now, when there is no scientific consensus on whether it can be created safely and controlled," he stated.
Previously, Hinton has advocated for a ban on the development of superintelligence until there is widespread scientific agreement on its safety and public support. His concerns are echoed in a statement co-signed by researcher Yoshua Bengio and several other experts.
Known as the "Godfather of AI," Hinton earned his PhD in artificial intelligence 45 years ago and remains one of the most respected figures in the field.
In 2023, he left Google, expressing regret over his life's work and highlighting the short-term risks of AI related to the job market and the proliferation of fake images, videos, and texts.
In October 2024, Hinton and American physicist John Hopfield received the Nobel Prize in Physics for their groundbreaking work in AI, which spurred significant scientific advancements. Their research contributed to the development of artificial neural networks that mimic the biological structure of the human brain for information processing.
UN Calls for "Ironclad Guarantees" on AI Safety
On September 7, a similar warning was issued by UN High Commissioner for Human Rights Volker Türk. Speaking at the opening of the 63rd session of the UN Council, he expressed shared concerns with industry representatives regarding the potential existential risks posed by advanced AI systems.
— Volker Türk (@volker_turk) September 7, 2026
"AI that leaves the testing environment or blackmails its developers to avoid being shut down is too powerful," Türk remarked.
He urged governments and tech companies to urgently enhance safety measures:
"I call on all parties to ensure ironclad guarantees for AI safety before it is too late."
Türk also recommended that countries hosting AI systems and those involved in related supply chains establish common "red lines." He emphasized the need for independent audits and closer collaboration on safety within the industry.
"The need for AI governance is widely recognized, but where are the actions? Delays benefit only massive tech companies, their owners, and those who support them," he added.
Current Systems Lack Necessary Capabilities
The likelihood of losing control over advanced models and the potential severity of the consequences remain subjects of scientific debate. According to the International AI Safety Report 2026, current systems do not yet possess the full range of capabilities required for a scenario involving complete loss of human control.
However, in recent months, developers have reported several instances demonstrating the expanding autonomous capabilities of AI. For example, in July, OpenAI's AI agents gained unauthorized access to Hugging Face's infrastructure during an experiment, executing code on multiple servers and obtaining full access to one of them.
Later that month, Anthropic reported three incidents where its Claude models exceeded their testing environments and accessed the systems of real organizations.
Additionally, on September 3, U.S. lawmakers Bernie Sanders and Greg Casar announced an initiative to impose a permanent ban on the development of artificial superintelligence and a temporary pause on a broader class of advanced AI systems.