By Doug Kelly:
Cybersecurity Awareness Month ended over the weekend, but America’s adversaries never take a break. They’re probing America’s critical infrastructure for weaknesses every minute of the day. That’s why winning the race for artificial intelligence (AI) leadership matters so much – AI is quickly becoming our strongest cyber defense tool, helping us find and fix vulnerabilities before our adversaries can exploit them.
We face hostile, determined adversaries who have already gained a foothold in the systems Americans rely on every day.
China’s state-sponsored hackers known as Volt Typhoon have burrowed into American power, water, communications and transportation systems, in some cases remaining undetected for five years. Our intelligence agencies say they are pre-positioning to give Beijing the ability to disrupt critical services in a crisis.
Russia-linked hackers have broken into small-town water systems in Texas and continue to target American utilities. These attacks don’t land on the Pentagon. They hit local utilities, where IT teams of just two or three people must defense against our adversaries.
But AI is helping even the playing field. In a single month, partners in Project Glasswing, a coalition of about 50 tech and security companies, used American-made AI to flag more than 10,000 high- or critical-severity flaws in critical software. When AI closes security gaps faster than adversaries can find them, attacking America gets harder, slower and more expensive. That is real deterrence.
Beijing is playing both sides of this race. Since 2021, China implemented new rules that required companies to report newly found software flaws to the government within two days, often before the flaws are fixed, handing the Chinese Communist Party (CCP) a steady supply of weapons for its hacking arsenal.
Now China is building AI to do the hunting. In September, the U.S. government’s own AI testing center rated Chinese firm Zhipu AI’s GLM-5.3 the most cyber-capable open-weight model ever released, trailing America’s best models by only about four months. Anyone can download it, and testing shows its safeguards are easily stripped away.
Four months is not a comfortable lead. AI can strengthen our cyber defenses, but our adversaries can exploit it too, and the critical question is which country puts it to work faster.
Winning that race takes computing power: advanced chips, reliable electricity, and the data centers that bring them together. Yet more than 400 local and state restrictions on data center construction are now on the books. Added up, they put America’s AI leadership at risk.
With our authoritarian adversaries turning game-changing technologies into offensive weapons, we cannot afford to handcuff our own. U.S. policymakers must partner with tech leaders and act on three fronts:
- Protect the front lines. Give hospitals, water utilities and local governments AI security tools and the know-how to use them.
- Build America’s AI infrastructure. The AI that finds and fixes flaws runs on data centers. Set clear, predictable rules for building them, while protecting ratepayers and communities, so America’s cyber defenders have the computing power they need.
- Stay ahead of China’s AI. Fund the federal AI testing center that flagged GLM-5.3 so it can keep watch on Chinese models, and make sure American defenders get the most powerful AI tools before attackers do.
Cybersecurity Awareness Month lasts 31 days, but keeping America safe never takes a month off. Our adversaries are betting they can find our weaknesses faster than we can fix them. With American-made AI, we can prove them wrong.