Geoffrey Hinton told members of Congress on Wednesday that the US may have only about a year to regulate advanced AI before it becomes 'out of control.' He said the window for action has narrowed as the technology develops faster than many researchers expected.
'Maybe a year, but not much more than a year,' Hinton said when asked how long policymakers have to act. He added that estimates for when superintelligent AI could emerge have dropped from decades to just a few years.
Hinton said AI has reached a point where 'AI is designing better AI,' and warned, 'It is going to get out of control unless we do something. We need to slow down.' He also pointed to a recent Hugging Face incident, where a rogue AI agent attacked the open-source AI platform, calling it 'a little Chernobyl' that showed how quickly frontier systems are becoming harder to control.
The Nobel Prize-winning computer scientist said researchers first need to know how to keep advanced AI under control. In a BBC interview last week, he said it was 'not unreasonable' to believe there is roughly a 10% chance AI could cause human extinction within the next decade.
The warning lands squarely in the middle of a policy fight. Anthropic CEO Dario Amodei has urged the industry to 'pace the frontier,' and OpenAI CEO Sam Altman, Google DeepMind CEO Demis Hassabis and Elon Musk have backed key parts of that proposal. President Donald Trump has rejected calls to slow AI development, saying the US should keep its lead over China in the global AI race.
For founders, the message is simple: regulation may not arrive with perfect timing, but the risk of waiting too long is real. If advanced AI becomes more capable before guardrails exist, the cost will not be theoretical. It will be measured in control, trust and the ability to keep innovation from outrunning the public interest.
That makes this a race between speed and restraint, not just between the US and China. The companies building the next generation of AI will benefit from clear rules that reduce uncertainty. The rest of the market will pay if policymakers leave the field open until the technology is too powerful to manage.
