Recent malfunctions — including a global Microsoft outage, an FAA grounding of air travel and a consumer banking glitch at Chase — were a mild preview of what experts expect, Axios reported.
Most resulted from companies and government agencies rushing to secure systems against future large-scale cyberattacks, a pattern that will persist. Future disruptions will increasingly be actual attacks carried out by U.S. rivals or smaller bad actors using powerful AI to spot and exploit vulnerabilities. “Recognize that this is coming.”
Double-edged sword of frontier AI
America’s frontier AI labs — Anthropic, OpenAI, Google and others — built the most powerful AI for building software, Axios reported. That same capability makes AI equally great at finding and exploiting vulnerabilities, often called bugs or holes. Government and business operate mostly on pre-AI computer systems full of those vulnerabilities.
One exploited flaw can shut down an entire company or all air travel, as recently witnessed. Anthropic previously held back its Mythos AI model because of its cyber capabilities, Axios reported. U.S. intelligence agencies did the same. Most of government and American business have not.
China’s entrenched presence
The full number of compromised systems remains unknown.
We have learned that third parties who provide services to the firm were the target of a cyber attack. Based on our investigation so far, we believe the attackers accessed certain marketing data held by those vendors, which may include some customer names and email addresses.…
Cyber defense costs limit access for most organizations
Cyber defense is costly, Axios reported. Many government agencies and businesses cannot afford to safeguard their systems. The biggest companies outbid others for top-tier cyber protection; the best firms are largely booked.
Many businesses hesitate to spend millions on pure loss prevention unless convinced a threat is imminent. The Texas energy regulator sounded alarms to the state’s oil, gas and utility networks that they are a target. The EU warned governments about ARTEX after a 10-day cyberattack last month.
Leadership shuffle at the White House
Davies was the Pentagon’s chief information officer and has deeper subject-matter expertise than her predecessor. Even the White House’s Super Intelligence Force is scrambling to get its head around rising risks while appraising near-daily reports of possible breaches, Axios reported. Trump shows little AI expertise. Congress is no better.

Siloed experts expect inaction until catastrophe strikes
Officials with cyber expertise across different sectors fear the same outcome but rarely talk to people outside their own domain and party, Axios reported. There is no real mechanism or institution to pull them together. Most assume no serious preventive action will happen until a true mini-catastrophe hits, yet they all assume a mini-catastrophe is coming.
Reasons not to freak out
Everyone interviewed expects attacks, Axios reported. Few predict true catastrophe. A big gap exists between a bank going out for a day and a region going dark for a week. A society-shaking attack breaking pipelines or blacking out a region still probably requires a nation-state.
China, the primary actor capable of such an attack, pays the highest price if it does — an attack on the U.S. grid is an act of war inviting response against the assailant’s systems.
Anthropic previously held back its Mythos AI model because of its cyber capabilities, Axios reported. The company’s model class Claude Mythos, a less restricted version of Fable, is available only to approved US organizations with a cybersecurity and life sciences focus, reflecting a controlled-access approach to dual-use technology.
This restriction demonstrates Anthropic’s willingness to limit deployment of its most capable systems when they present elevated safety risks, particularly in domains involving vulnerability discovery and exploitation.
Potential solutions emerging
The government, using pressure or requirements, could bring AI companies together to help strengthen vulnerable American systems before fully focusing on building more powerful models, Axios reported. Top executives at leading AI companies say they would gladly comply if rivals did. These leaders feel post-election urgency to act before a public event makes citizens hate AI even more.
Jay Clayton, chair of the White House task force, is finishing a voluntary safety regime for big AI companies and is probably best positioned to pull AI CEOs together, according to Axios. He is well-versed in cyber and sees AI largely through a national security lens.
More on Recent
Keep reading