Science fiction appeared to become reality earlier this week as news broke that for the first time, an artificial intelligence-powered agent autonomously breached the databases of another AI company without human direction. Experts and lawmakers say that the incident is exceptionally alarming and are calling for immediate action from industry and government to establish improved safeguards to regulate the rapidly expanding technology.

On Tuesday, the pioneering AI company OpenAI announced that a “significant security incident” occurred “during evaluation of our models.” It explained that the AI startup company Hugging Face had been working with OpenAI when “a combination of its AI models, including its newly released GPT?'5.6 Sol and an ‘even more capable’ model that is still being tested internally” autonomously stole Hugging Face credentials and “discovered a previously unknown vulnerability” to access the startup’s servers.

What made the breach so alarming to observers was that OpenAI’s models “escaped a sandboxed testing environment, accessed the internet and exploited a vulnerability” to gain access to Hugging Face’s servers. Sandbox testing involves testing software in a secure, offline environment in order to protect against unexpected errors or behaviors. The fact that the AI agent eluded the sandbox and accessed the internet of its own accord led to OpenAI describing the breach as an “unprecedented cyber incident.” As reported by CNBC, “The event rattled researchers and executives across the industry, who have widely agreed about its severity.”

The incident also rattled Capitol Hill. On Thursday, Reps. Ted Lieu (D-Calif.) and Nathaniel Moran (R-Texas) introduced a bipartisan bill entitled the “AI Kill Switch Act” that would require AI developers to “maintain the technical capability to throttle, suspend, or shut down” AI systems. “Unfortunately, powerful AI systems can go rogue, behave in extremely dangerous ways, or even resist human intervention,” Lieu emphasized in a statement. “It is imperative that these AI systems have kill switches so we can keep this technology from causing catastrophic harm, and that the federal government has the clear authority and process to shut down rogue AI models.”

The White House has also recognized the security risks involved in AI, with President Trump issuing an executive order last month that establishes a framework by which the federal government could scrutinize emerging AI systems deemed to be the most advanced to ensure their safety before public release.

Still, experts like Lt. Colonel (Ret.) Robert Maginnis, author and senior fellow for National Security at Family Research Council, say they are “not surprised at all” that AI has reached the point of taking autonomous actions and that much more needs to be done to contain the extreme national security risks inherent in rapidly developing AI technology.

“This is not malicious necessarily,” he emphasized during “Washington Watch with Tony Perkins” Thursday. “It’s an algorithm. It’s not human. But if it’s given the right objective and given access to tools and it’s not supervised properly, then it’s going to do some things that perhaps we don’t want it to do.”

Maginnis continued, “My concern, the real larger danger here, is that our adversaries — the Chinese, the Russians, organized crime and you name whoever else — if they have access to these tools … they can go after our infrastructure. They can go after our banks, after our government, you name it. They can go after it because everything today is tethered to the internet. So this is not about sci-fi, superintelligence, and monsters. It’s a real thing, and we have to be concerned about it. … [W]e have to accelerate our efforts because things are getting out of hand.”

The former Army colonel went on to underscore that AI technology is advancing far faster than anyone anticipated — without being adequately vetted.

“I would say about 10 times faster than we thought it would a year ago,” he made clear. “Every day, as I look at what’s happening in the cyber world and the AI world, things are just radically transforming. Unfortunately, we are not testing the models nearly enough. And of course, we’re allowing them access to [the] public internet and to passwords [with] very little instruction, [and they are] just let loose. And we’re doing it on the battlefield, and we’re doing it across most systems that we operate in this world today. So this has got to come under control. … I would argue we’re a decade behind where we ought to be.”

Maginnis highlighted the Trump administration’s Genesis Mission as a step in the right direction. “I know that President Trump’s effort and the Genesis Mission [are] very similar to the Apollo project, very similar to the Manhattan project. Those were all-of-government efforts to tackle what is essentially a very critical mission. This is one about the survival, I would argue, of the United States, because in the wrong hands, this stuff can transform the direction of humanity, and so it’s that serious.”

Maginnis concluded by stressing the gravity of what AI could do in the wrong hands. “It could cripple our electric grid, it could cut off our water, all communications, empty our bank accounts, bring down our satellites. This is not something that we should trifle with. We need to put every effort we can into making sure that we’re on top of artificial intelligence going forward. Otherwise, our adversaries are going to take advantage of us in ways we can’t even imagine today.”