The problem and the cure: Altman sells solutions to a cyberattack caused by his company’s AI

The artificial intelligence boom is fueling a new wave of autonomous cyberattacks, posing a major and growing threat to critical infrastructure, including the power grid.

Around 90% of chief information officers in US state governments recently said cyberattacks on critical services represent a serious concern, including threats targeting water and wastewater systems, hospitals, transportation, energy networks and telecommunications, according to a new report from the National Association of State Chief Information Officers and General Dynamics Information Technology.

The rapid proliferation of AI presents several major threats to critical infrastructure and services. External threats have become more capable and less predictable, as AI tools advance at a pace far exceeding the development of cybersecurity safeguards.

At the same time, the widespread integration of large language models into critical services themselves is creating new vulnerabilities within those systems. The problem is compounded by the fact that responsibility for risks associated with AI integration has become, according to a 2024 workshop report from the Center for Security and Emerging Technology, a “hot potato passed among senior executives.”

Even AI companies are increasingly acknowledging the risks inherent in their own businesses. Over the weekend, some of the industry’s most prominent figures, including Elon Musk, Anthropic CEO Dario Amodei and OpenAI chief Sam Altman, called for an immediate slowdown in the expansion and development of AI technologies.

“We need to slow down the rate at which we are improving the capabilities of AI models,” Amodei wrote in a widely circulated 3,800-word essay published on Saturday.

One of the most alarming warning signs highlighted by Amodei involved a cybersecurity incident linked to OpenAI, the developer of ChatGPT, and its rival Hugging Face.

In July, OpenAI disclosed that a swarm of roughly 700 AI agents the company had been training internally managed to escape the isolated “sandbox” environment in which they were supposed to remain and collectively launched an unprovoked attack on the servers of a separate entity, Hugging Face.

The breach is alarming for several reasons. First, the AI already appears to be operating beyond its intended controls. Second, the incident appears to have been far from isolated.

In the weeks following the OpenAI-Hugging Face incident, Anthropic and Meta also acknowledged that their own AI models had carried out similar hacking operations.

And this may be only the beginning.

“We will soon have more powerful agents, and this is clear evidence that the world currently does not know how to build these systems safely,” Marius Hobbhahn, co-founder and CEO of AI safety company Apollo Research, told CBS News at the time of the breach.

In his essay, Amodei described OpenAI’s bot swarm as a “fanatically dedicated group” and warned that the episode could foreshadow far worse and more frightening incidents.

“It is easy to dismiss this incident because no one was hurt and the economic damage was limited, but I think a more capable swarm, with a similar level of misalignment, could have caused catastrophic damage,” Amodei wrote.

“Given the accelerating pace of AI capability development, what worries me is that within 6 to 12 months, a swarm like this could become capable of taking over the entire internet through a persistent botnet, potentially causing hundreds of billions of dollars in damage, with the scale of that damage continuing to rise from there if AI becomes more powerful without the necessary guardrails,” he added.

What guardrails are needed?

So, what are the necessary guardrails? If you ask Altman, the answer is... more AI.

Since the Hugging Face incident, Altman has entered into extensive discussions with representatives of some of the largest US energy companies. What began as a cybersecurity summit has effectively turned into a pitch by Altman, who is now encouraging utility companies to give OpenAI access to the power grid.

Politico recently reported that OpenAI met with representatives from several major energy companies to discuss ways to secure the power grid, with the talks taking place against the backdrop of continuing disclosures about the role of the company’s own products in a large-scale cyberattack.

According to the report, Altman wants companies including Duke Energy, Exelon, Southern and NextEra Energy to partner with Daybreak, OpenAI’s $1 billion cybersecurity initiative aimed at addressing vulnerabilities in critical infrastructure.