AI ‘Kill Switch’ Could Go as Far as Shutting Down the Internet

You may have heard a lot of buzz lately about the so-called "AI kill switch." What is this exactly, and would it effectively guard against the dangers of artificial intelligence (AI)? This is what you need to know.

Understanding What a Kill Switch Would Do

The term AI kill switch is a general phrase applying to any type of guardrail that would work to mitigate the ability of AI agents to take dangerous actions. Examples of these unwanted actions include mass software and hardware interference that could trigger negative consequences.

The topic of developing an AI kill switch has taken center stage in recent weeks as the threat of what AI is capable of is becoming more apparent. Several members of Congress on both sides of the aisle have introduced kill switch bills that would mandate that developers be able to suspend or shut down their products in an instant.

Anthropic co-founder Jack Clark supports the idea of mandatory "kill switches" for this type of technology, recently telling the media that “I think that’s the kind of thing society is going to want to know and might want to eventually pass rules around.”  

According to Michael Vermeer, a physical scientist at the nonpartisan research organization RAND, building an AI kill switch "could be trivially easy" on a local level. Vermeer said that the switch "could be literally pulling a plug, cutting the power, blocking internet access, something like that."

The concept gets a little more complicated on a global scale, potentially having the power to shut down the internet completely. The goal of this type of roadblock is to be able to remove an AI agent's access to a system that it should not be allowed to access.

For example, a kill switch could turn off the power to servers or build restraints into the electronic circuits that support the AI models. The switch could also help humans to prevent negative outcomes before they are unleashed.

One of the biggest complications with the concept of a kill switch is that they are not universal. In other words, pulling the switch to shut down one agent or company would not prevent other models from continuing the potentially dangerous actions.

In addition, while a kill switch might shut down one rogue agent, the harness could transition to another bad actor to carry out the actions. For instance, if a kill switch were implemented on OpenAI, the connection could simply switch to Anthropic. These agents could even switch to a model from abroad, potentially proving to be even more uncontrollable. In order to guarantee that the AI switch would be effective, it would need to be able to shut down the entire internet on a global scale.

Why Are Some Lawmakers Pushing for a Kill Switch

So why are some lawmakers concerned about the possibility of the dangers unleashed by AI? Tech companies note that an agent has "gone rogue" when it inadvertently takes actions not supported by humans. This could translate to major cybersecurity breaches or life-ending scenarios.

Vermeer said that the big question is whether a kill switch would be triggered by a company if an agent went rogue. Some tech leaders may be hesitant to pull the switch because it could put the company at risk of liability or customer dissatisfaction.

“If the people who are responsible for using the kill switch are not liable for damages if they don't use it, it's very unlikely, or it wouldn’t necessarily be rational, for them to ever use it,” Vermeer warned.

Reps. Ted Lieu and Nathaniel Moran first introduced the bipartisan “AI Kill Switch Act” in July. This bill would give the federal government the power to order a company to activate its kill switch. In addition to the bill in the House, Sen. John Kennedy also confirmed that he plans to introduce similar legislation in the Senate.

President Donald Trump does not seem to think that a kill switch is necessary. The president recently posted on Truth Social, "The only control or 'guardrails' that AI needs is a STRONG AND SMART (High IQ!) PRESIDENT, and the U.S.A. has that, in spades!"

Other industry experts believe that the technology could be better used to address the buggy and insecure code that is all over the internet. Rather than looking at AI as the problem, some people believe that we should lean on the power of AI to detect the bugs and fix them so that the world is less vulnerable to rogue agents.

It could be years before the kill switches would even come to fruition. However, many tech companies are already working on their own guardrails. For instance, OpenAI announced last week that it is going to begin routing allegations of misuse to an internal "Safety Advisory Group" for further escalation.

Looking for stories that inform and engage? From breaking headlines to fresh perspectives, WaveNewsToday has more to explore. Ride the wave of what’s next.