Safety

AI-generated text

AI Raises Cybersecurity Risks for Critical Infrastructure

Government, business and AI leaders interviewed for the source article expect increasingly sophisticated cyberattacks and disruptive outages over the next 18 months.

AI Raises Cybersecurity Risks for Critical Infrastructure

More than 50 leaders from government, artificial intelligence, business and cybersecurity interviewed for the source article expect attacks on companies and critical infrastructure to become constant, more sophisticated and increasingly disruptive over the next 18 months. The interviewees included both advocates of rapid AI development and its sharpest critics.

Several believe that by 2027 cyber risk could become a routine condition of daily life, with people and businesses having to navigate frequent and occasionally frightening outages. The article presents this as the leaders’ expectation rather than a certain forecast.

The article points to disruptions during the preceding two weeks, including a global Microsoft outage, a halt to air travel ordered by the U.S. Federal Aviation Administration (FAA), and a glitch affecting Chase consumer banking services. It describes them as a mild preview of potential future disruptions. According to the report, most of the malfunctions appeared to result from companies and government agencies trying to secure their systems against future attacks, rather than from actual cyberattacks.

In an assessment published that month, hedge fund Bridgewater Associates said it was managing money on the basis of a one-in-three chance of an economically destabilizing attack within a year. Horacio Rozanski, chairman and CEO of Booz Allen, warned that autonomous attacking AI agents could operate at extraordinary speed once their capabilities mature, and urged organizations to take as many precautions as possible now.

AI as both a defense and an attack tool

Advanced AI systems from Anthropic, OpenAI and Google are particularly useful for building software. Those same capabilities can also help identify and exploit programming errors and security gaps. Most government and corporate systems were built before the AI era and may contain numerous vulnerabilities. Exploiting a single weakness, or making an error during a good-faith repair, can cause a major outage.

According to the source article, Anthropic previously held back its Mythos model because of concerns about its cyber capabilities. Since then, some global companies, including JPMorgan Chase, and U.S. intelligence agencies have used Mythos and other technologies to find and repair vulnerabilities. Most American government bodies and businesses, however, have not taken similar steps, the article says.

Costs and shortages of expertise also hinder defense. The largest companies can outbid others for leading providers, whose capacity is largely booked, while other organizations cannot afford adequate protection or postpone major spending.

Groups of autonomous AI agents could add to the risk. On November 13, the U.S. Federal Bureau of Investigation (FBI) warned about ARTEX, free, open-source software that can use AI agents to carry out multiple stages of an attack at machine speed. The Texas energy regulator warned the state’s oil, gas and utility networks that they could be targets; the European Union also warned governments about ARTEX after a 10-day cyberattack the previous month.

Old systems and known intruders

The article says Volt Typhoon, a Beijing-sponsored hacking group, has been lurking in U.S. critical infrastructure for more than five years, including water, energy, transportation and communications networks. A related group, Salt Typhoon, broke into major U.S. telephone carriers and obtained call records and data related to wiretaps. The total number of affected systems is unknown.

Even the best-case scenario described by the interviewees involves regular, unpredictable outages and smaller scares, such as losing access to banking services or air travel for a day. Broad efforts to upgrade systems can themselves produce errors, while U.S. adversaries and smaller attackers may use increasingly powerful AI tools to exploit weaknesses.

The article says that until recently the White House and Congress focused more on safety regulation and data centers than on immediate cyber risks. The White House Super Intelligence Force (SIF) is now assessing the risks and near-daily reports of possible breaches. President Donald Trump removed national cyber director Sean Cairncross and appointed Kirsten Davies, the Pentagon’s former chief information officer, to replace him; the article says Davies has deeper subject-matter expertise than her predecessor.

The source article also identifies shortages of government expertise and staff, as well as weak coordination between institutions, as problems. Experts across sectors and levels of government share similar concerns but rarely consult one another, and there is no effective mechanism to bring them together. Several fear that substantial preventive action will begin only after a serious incident.

What could reduce the danger?

The interviewees expect attacks, but few predict a society-wide catastrophe. The article distinguishes between a bank outage lasting one day and a week-long blackout across a region. An attack of the latter kind, or one damaging pipelines, would probably still require the capabilities of a state actor and could invite a severe response, potentially including war. The fragmented U.S. electricity system, with thousands of utilities, also makes it more likely that a successful attack’s effects would remain regional.

One possible response proposed by the article is for the U.S. government to encourage or require AI companies to work together to strengthen vulnerable systems. Executives interviewed said their companies would be willing to participate if their rivals did the same. Jay Clayton, chair of the White House task force, is working on a voluntary safety framework for major AI companies and, the article suggests, could bring their chief executives together.

In the private sector, JPMorgan Chase chairman and CEO Jamie Dimon is leading the Alliance for Critical Infrastructure (ACI). The group brings together major finance, energy, technology, telecommunications and airline companies to strengthen resilience, reduce systemic risk and support continuity across the economy.

The source article concludes that the leaders interviewed collectively consider an actual attack more likely than timely cooperation to prevent one.