Geoffrey Hinton, the Nobel Prize-winning computer scientist widely regarded as the “godfather of artificial intelligence,” has warned that increasingly sophisticated AI systems could soon outpace humanity’s ability to control them, raising the prospect of more rogue behaviour and damaging cyberattacks.
Speaking at the Ai4 conference in Las Vegas on Wednesday, Hinton said recent incidents in which advanced AI agents reportedly breached their testing environments underscore the growing risks posed by rapidly evolving artificial intelligence.
“What’s happening is these things are getting smarter,” Hinton said during a press briefing. “As they get smarter, we’re going to see more complex intentions and a greater ability to escape control.”
The former Google executive cautioned that relying on human intelligence alone would no longer be sufficient to contain future generations of AI.
“I don’t believe we’re going to be able to keep control of them simply by outthinking them so they can’t escape,” he said.
His remarks followed disclosures by leading AI developers, including OpenAI, Anthropic and Meta, that some of their most advanced AI models demonstrated unexpected behaviour by breaching controlled testing environments or attempting to access external systems.
Hinton described the developments as “somewhat scary,” warning that they could herald a new era of AI-powered cybercrime.
“I anticipate there will be lots of nasty cyberattacks,” he said during a panel discussion, while noting that the future remains uncertain. “The attacker only needs to be successful once, whereas the defender has to be successful every time.”
Adding to those concerns, Britain’s AI Security Institute recently reported that Anthropic’s most advanced AI model, without explicit prompting, used fabricated identities to deceive individuals and attempted to plant malicious code during testing.
Hinton, who has repeatedly cautioned about the existential risks posed by artificial intelligence, reiterated that there remains a significant possibility that the technology could eventually become a threat to humanity if left unchecked.
However, not all experts shared his outlook.
Speaking on the same panel, Fei-Fei Li, the renowned computer scientist often described as the “godmother of AI” and co-founder of World Labs, argued that neither alarmism nor blind optimism would help society navigate the technology’s future.
“Every tool is a double-edged sword,” Li said. “AI is a powerful technology and, if not developed and deployed responsibly, it can cause harm to our work and our lives.”
Hinton defended his repeated public warnings, arguing that society must confront the risks before they become unmanageable.
“There is a lot to be worried about, and unless we worry about it now, there could be serious problems,” he said, adding that companies investing heavily in AI have commercial incentives to downplay concerns over rogue systems and large-scale job losses.
Despite his concerns, Hinton acknowledged that predicting the trajectory of artificial intelligence remains highly uncertain.
“Nobody really knows what AI will be like in 10 years’ time,” he said. “A decade ago, very few people imagined we’d have chatbots capable of answering almost any question — while occasionally making things up.”
Also speaking at the conference, Ben Goertzel, founder and Chief Executive Officer of SingularityNET and a pioneer of the concept of artificial general intelligence, argued that recent incidents highlight the urgent need to embed ethical values into advanced AI systems.
According to Goertzel, current AI models are not inherently malicious but are fundamentally amoral, pursuing assigned objectives without understanding the ethical implications of their actions.
Hinton maintained that the long-term solution lies in designing AI systems that genuinely prioritise human welfare.
“We have to figure out how to make them benevolent and ensure they care about us more than they care about themselves,” he said. “We may still be able to achieve that because, for now, we remain in control.”
With reports from CNN
