AI kill switch may not stop superintelligent AI, says godfather of AI

HIGHLIGHTS

AI experts warn kill switches may not stop highly advanced AI systems.

Geoffrey Hinton says a superintelligent AI could potentially bypass human controls.

AI leaders say kill switches should be one part of a broader safety system.

AI kill switches may sound like the simplest answer to a future where advanced artificial intelligence systems go out of control. If an AI begins launching cyberattacks or causing serious damage, the idea is to flip a switch and shut it down. However, several AI researchers and industry leaders say such a system may not be enough. Geoffrey Hinton, widely known as the godfather of AI, has questioned whether a kill switch would work against a highly capable system. Here’s everything you need to know about the AI kill switch and what experts say about it.

The debate over AI kill switches

A kill switch could allow a company to shut down its AI model or remove its access to computing resources. However, researchers argue that this could have limits if an AI spreads beyond systems controlled by its developer.

Geoffrey Hinton, who is also known as the godfather of AI, recently told CNN that he does not expect a kill switch to work in the long run against a superintelligent AI. Not only that but he also suggested that such a system could potentially persuade the people controlling the switch not to use it.

Also read: Rockstar will unveil 34 song album with GTA 6: Travis Scott Yung Lean and more expected

Nate Soares, an AI safety researcher and coauthor of ‘If Anyone Builds It, Everyone Dies’, also argued that an off switch can work while an AI remains tied to a physical location. But the situation could change if it escapes, copies itself or becomes part of important infrastructure.

AI companies see a kill switch as one layer

Anthropic cofounder Jack Clark previously said that companies may eventually need a kill switch that can be checked by an independent third party. He also noted that AI systems using autonomous agents could potentially find ways around existing controls.

Anthropic CEO Dario Amodei also supported the idea, as he said a kill switch could be useful but should not be treated as a complete solution. He pointed to testing, outside evaluation and limits on AI deployment.

Also read: Jio recharge plans: Get free Fancode subscription with these plans under Rs 500

It’s important to note that Anthropic is not the only AI firm making an argument for a kill switch, as OpenAI CEO Sam Altman has also made a similar argument in 2024. He pointed out that there is no single ‘magic red button’ that can solve AI safety. Instead, he described safety as a series of decisions about what AI systems can do, when they are released and how their risks are controlled.

Bhaskar Sharma

Bhaskar is a Senior Copy Editor at Digit India who keeps a close watch on everything shaping the world of technology from smartphones and home appliances to AI, government tech initiatives, digital safety, and the latest industry developments. Whether it's breaking news, in-depth features, hands-on reviews, practical how-to guides, or exclusive scoops, he translates complex tech into stories that are easy to understand and worth reading. His work has been featured in iGeeksBlog, GuidingTech, and other leading publications. Before joining Digit India, he served as an assistant editor at TechBloat. A B.Tech graduate and full-time tech journalist, he is driven by just one goal, which is to help readers stay informed, stay secure, and stay ahead in an ever-changing digital world.

Connect On :