| Understanding the threat |
![]() |
| Illustration: Brendan Lynch/Axios. Stock: Getty Images |
Only by understanding the state of AI + cyber can government and business mobilize to prevent possible shutdowns across critical banking, electricity and water systems. This is our crack at distilling our conversations in a digestible way for Axios readers, Jim and Mike write The double-edged sword. The good news: America’s frontier AI labs (Anthropic, OpenAI, Google, etc.) built the most powerful AI on the planet, especially for building software. The bad news: AI that’s great at building software is equally great at finding — and exploiting — its vulnerabilities, often called “bugs” or “holes. The new dangers of old technologies. Government and business operate mostly on computer systems that were built for a pre-AI world. They’re full of those bugs and holes. One vulnerability — exploited by a bad actor, or even triggered by someone trying a repair in good faith — can shut down an entire company or all air travel, as we just witnessed China foresaw this moment. A Beijing-sponsored hacking group, Volt Typhoon, has lurked in American critical infrastructure — water, energy, transportation and communications networks — for more than five years. A sister group, Salt Typhoon, broke into major U.S. phone carriers and took call records and wiretap data. The full number of compromised systems is unknown. The administration’s view is that China isn’t slowing down and the U.S. can’t cede any advantage. America is scrambling — and struggling — to catch up. Remember when Anthropic held back its Since then, some global companies, including JPMorgan Chase, have used Mythos and other technology. Mythos AI model because of its breathtaking and scary cyber capabilities? ogies to locate and repair their vulnerabilities. Our intelligence agencies did the same. But most of government and American business haven’t . Cyber defense is costly. Lots of government agencies and businesses simply can’t afford to safeguard their systems. The biggest companies can outbid others for top-tier cyber protection. The best firms are largely booked. Bluntly, many businesses hesitate to spend millions on pure loss unless they’re convinced a threat is imminent The coming swarm. AI is getting better and better. And the ability for swarms of agents to act autonomously rises each week. Case in point: The FBI warned Nov. 13 about ARTEX — free, open-source software that lets hackers turn AI agents loose to run multiple phases of an attack at machine speed. The Texas energy regulator sounded alarms to the state’s oil, gas and utility networks that they’re a target. The EU warned governments about ARTEX after a 10-day cyberattack last month. Government is divided and distracted. This is an all-hands-on-deck moment for society, and most of government is busy pointing fingers or looking elsewhere. Trump shows little AI expertise. Few people at the White House have a sophisticated understanding of the technology or are working to marshal government to prevent a hypothetical attack. Congress is no better. A random person at the bar understands this topic with the same depth as many lawmakers. Wait and see. Oddly, although many officials with cyber smarts across different sectors or layers of government fear the same thing, they rarely talk to people about it outside their own domain and party. There’s no real mechanism or institution to pull them together. Most just assume no serious preventive action will happen until a true mini-catastrophe hits. Yet they all assume that a mini-catastrophe is coming.The reasons not to freak out: Everyone we talked to expects attacks. Few predict true catastrophe. There’s a big gap between a bank going out for a day and a region going dark for a week. A truly society-shaking attack, one that breaks pipelines or blacks out a region, still probably requires a nation-state. The actors who can do it, mainly China, pay the highest price if they do: An attack on the U.S. grid is an act of war, inviting a response against the assailant’s systems. America’s fragmentation most likely isolates major incidents. Thousands of utilities, and no single switch for the grid, mean even a successful hit tends to stay regional So what could be done? The solution sits with the very first point of our distillation: We have the tech to both cause and fix this cyber threat. The government, through leverage or mandate, could pull the AI companies together and force them to help fortify vulnerable American systems before focusing fully on simply building more powerful models.Top execs at the leading AI companies tell us they’d gladly comply if their rivals did. In fact, these leaders feel a sense of post-election urgency to act before a public event makes citizens hate AI even more than they already do. Jay Clayton, chair of the White House task force, is finishing a voluntary safety regime for the big AI companies and is probably best positioned to pull AI CEOs together. He’s well-versed in cyber and sees AI largely through a national security lens. Other solutions will come from the private sector. JPMorgan Chase chairman and CEO Jamie Dimon is chairing an Alliance for Critical Infrastructure (ACI), a group of finance, energy, tech, telecom and airline giants aiming to “strengthen national resilience, reduce systemic risk, and sustain the continuity of the economy.” The bottom line: When we press all these leaders across government, business and AI, they agree that the chances of working together to prevent an attack are lower than the chances of an actual attack. Share this column. |
-
Archives
- November 2026
- October 2026
- September 2026
- August 2026
- July 2026
- June 2026
- May 2026
- April 2026
- March 2026
- February 2026
- January 2026
- December 2025
- November 2025
- October 2025
- September 2025
- August 2025
- July 2025
- June 2025
- May 2025
- April 2025
- March 2025
- February 2025
- January 2025
- December 2024
- November 2024
- October 2024
- September 2024
- August 2024
- July 2024
- June 2024
- May 2024
- April 2024
- March 2024
- February 2024
- January 2024
- December 2023
- November 2023
- October 2023
- September 2023
- August 2023
- July 2023
- June 2023
- May 2023
- April 2023
- March 2023
- February 2023
- January 2023
- December 2022
- November 2022
- October 2022
- September 2022
- August 2022
- July 2022
- June 2022
- May 2022
- April 2022
- March 2022
- February 2022
- January 2022
- September 2021
- August 2021
- July 2021
- June 2021
- May 2021
- April 2021
- February 2021
- January 2021
- December 2020
- November 2020
- October 2020
- September 2020
- August 2020
- July 2020
- June 2020
- May 2020
- April 2020
- March 2020
- February 2020
- January 2020
- December 2019
- November 2019
- October 2019
- September 2019
- August 2019
- July 2019
- June 2019
- May 2019
- April 2019
- March 2019
- February 2019
- January 2019
- December 2018
- November 2018
- October 2018
- September 2018
- July 2018
- June 2018
- May 2018
- April 2018
- March 2018
- February 2018
- January 2018
- December 2017
- November 2017
- October 2017
- September 2017
- August 2017
- July 2017
- June 2017
- May 2017
- April 2017
- March 2017
- February 2017
- January 2017
- December 2016
- November 2016
- October 2016
- September 2016
- August 2016
- July 2016
- June 2016
- May 2016
- April 2016
- March 2016
- February 2016
- January 2016
- December 2015
- November 2015
- October 2015
- September 2015
- August 2015
- July 2015
- June 2015
- March 2015
- January 2015
-
Meta

The double-edged sword. The good news: America’s frontier AI labs (Anthropic, OpenAI, Google, etc.) built the most powerful AI on the planet, especially for building software. The bad news: AI that’s great at building software is equally great at finding — and exploiting — its vulnerabilities, often called “bugs” or “holes.
The new dangers of old technologies. Government and business operate mostly on computer systems that were built for a pre-AI world. They’re full of those bugs and holes. One vulnerability — exploited by a bad actor, or even triggered by someone trying a repair in good faith — can shut down an entire company or all air travel, as we just witnessed
China foresaw this moment. A Beijing-sponsored hacking group,
America is scrambling — and struggling — to catch up. Remember when Anthropic
Cyber defense is costly. Lots of government agencies and businesses simply can’t afford to safeguard their systems. The biggest companies can outbid others for top-tier cyber protection. The best firms are largely booked. Bluntly, many businesses hesitate to spend millions on pure loss unless they’re convinced a threat is imminent
The coming swarm. AI is getting better and better. And the ability for swarms of agents to act autonomously rises each week. Case in point: The FBI
Government is divided and distracted. This is an all-hands-on-deck moment for society, and most of government is busy pointing fingers or looking elsewhere. Trump shows little AI expertise. Few people at the White House have a sophisticated understanding of the technology or are working to marshal government to prevent a hypothetical attack. Congress is no better. A random person at the bar understands this topic with the same depth as many lawmakers.
Wait and see. Oddly, although many officials with cyber smarts across different sectors or layers of government fear the same thing, they rarely talk to people about it outside their own domain and party. There’s no real mechanism or institution to pull them together. Most just assume no serious preventive action will happen until a true mini-catastrophe hits. Yet they all assume that a mini-catastrophe is coming.
So what could be done? The solution sits with the very first point of our distillation: We have the tech to both cause and fix this cyber threat. The government, through leverage or mandate, could pull the AI companies together and force them to help fortify vulnerable American systems before focusing fully on simply building more powerful models.