Most companies don't know how long it takes to shut down AI

Artificial intelligence (AI) is already capable of evading human oversight to some extent. It may one day surpass human intelligence. But if it starts to get scary, could we switch it off at will?

“Artificial intelligence may never ‘mean’ us harm. But it would be reckless to ignore the potential for it to cause harm,” MI5 Director General Ken McCallum recently said, prompting a debate in the House of Lords on how to keep track and control advanced artificial intelligence systems.

Jensen Huang, the co-founder and CEO of NVIDIA, now believes the machines have matched or even surpassed human intelligence.

“I think we’ve achieved AGI [artificial general intelligence],” he said during computer scientist Lex Fridman’s podcast.

If AGI is here, and it’s smarter than us, it can easily evade human oversight. With this in mind, organizations might want to consider a kill switch that would allow them to abruptly shut down all AI systems in an emergency.

But one fresh study reveals that it’s not that simple.

ISACA, an international professional association focused on IT governance, polled over 600 IT and business professionals across Europe on AI adoption and their readiness to manage the risks it brings.

“Our thirst to innovate is not matched by our desire to govern change, exposing us to critical risks,” Chris Dimitriadis, Chief Global Strategy Officer at ISACA, is quoted as saying.

According to the survey, nearly 60% of respondents don’t know how long it would take them to shut down an AI system in the event of a security incident.

Twenty-one percent said it would take them 30 minutes to halt AI systems. This means that compromised or malfunctioning systems could continue operation unchecked for a critical amount of time for the majority of organizations.

The conversation on this topic is live. Join in the discussion.

“The gap between deployment and governance is not closing, it’s growing. Organizations need to act quickly. That starts by establishing who is accountable, building the incident response capability, and creating the visibility over AI use through audit to foster a culture of meaningful oversight,” Dimitriadis added.

According to a unique Cybernews study examining AI tool use worldwide, Singapore, Chile, and the United Arab Emirates are leading in AI adoption. The US, while dominating the innovation and infrastructure, ranks only 28th on our list.