Geoffrey Hinton, often called the “Godfather of AI,” warned in an International Business Times report published August 7 that the worst risks from AI may still be ahead as smarter models become harder for people to control. The article frames Hinton’s concern around increasingly capable AI systems and the control problem, meaning the difficulty of ensuring advanced systems behave as humans intend.

The report says Hinton voiced concern that more powerful models could become less manageable as their intelligence rises. The headline’s warning, “the worst is yet to come,” centers on the risk that future AI models may be harder to supervise, restrict or reliably direct than today’s systems.

The same IBTimes page places Hinton’s warning alongside other AI-risk and frontier-model stories, including claims that AI may soon improve without human involvement, that AI models are getting better at hacking, and that Google DeepMind is preparing for biosecurity risks. The page also lists business and infrastructure angles involving OpenAI, xAI, Palantir, Microsoft, Amazon, Hugging Face, Goldman Sachs and Lenovo, but Hinton’s control warning is the central item provided in the article text.

The article gives no benchmark scores, technical specifications or policy proposal details in the supplied text. Its verifiable facts are the publication date, Hinton’s stated concern about smarter AI models becoming harder to control, and the broader page context that groups his warning with other coverage of AI safety, frontier models, autonomous cyberattacks, biosecurity and AI infrastructure.

Geoffrey Hinton, often called the “Godfather of AI,” warned in an International Business Times report published August 7 that the worst risks from AI may still be ahead as smarter models become harder for people to control. The article frames Hinton’s concern around increasingly capable AI systems and the control problem, meaning the difficulty of ensuring advanced systems behave as humans intend.

The report says Hinton voiced concern that more powerful models could become less manageable as their intelligence rises. The headline’s warning, “the worst is yet to come,” centers on the risk that future AI models may be harder to supervise, restrict or reliably direct than today’s systems.

The same IBTimes page places Hinton’s warning alongside other AI-risk and frontier-model stories, including claims that AI may soon improve without human involvement, that AI models are getting better at hacking, and that Google DeepMind is preparing for biosecurity risks. The page also lists business and infrastructure angles involving OpenAI, xAI, Palantir, Microsoft, Amazon, Hugging Face, Goldman Sachs and Lenovo, but Hinton’s control warning is the central item provided in the article text.

The article gives no benchmark scores, technical specifications or policy proposal details in the supplied text. Its verifiable facts are the publication date, Hinton’s stated concern about smarter AI models becoming harder to control, and the broader page context that groups his warning with other coverage of AI safety, frontier models, autonomous cyberattacks, biosecurity and AI infrastructure.