National Cyber Director Sean Cairncross said Tuesday that the Trump administration wants AI models deployed across critical infrastructure sectors as quickly as possible to strengthen their defenses, CyberScoop reported. Speaking at an event hosted by the telecom industry group USTelecom, he said the government is already working with operators to embed AI systems in vital sectors.

What Happened

Cairncross framed rapid adoption as a security measure in its own right. According to CyberScoop, he said the aim is to find ways to protect the country's systems with the best available technology as fast as possible.

He also told executives where to focus. Rather than treating AI as a single tool, he said, company leaders need visibility into the AI agents operating on their systems and across their supply chains: who has authorized them, what they are permitted to reach and where in the network they run.

RelatedWhite House AI Task Force Plans Breach Warnings From Labs at Its First Meeting

Cairncross described what he called a "Mythos moment" in the spring, when the debut of a highly capable AI model brought the administration's efforts into focus and set off extensive coordination among federal agencies and with industry on security.

He argued that AI has mostly revealed long-standing weaknesses in cybersecurity rather than creating new kinds of threats. In that light, he said, replacing aging systems that no longer receive vendor support is a less visible but critical part of the strategy, and it will take financial incentives to get operators to do it.

USTelecom CEO Jonathan Spalter said AI models embedded in critical infrastructure have become dramatically more effective over the past six months, CyberScoop reported.

Why It Matters

The remarks show how the White House wants to handle AI security in sectors such as telecommunications, energy and water: through fast adoption and cooperation with operators rather than new mandates. CyberScoop noted that some lawmakers have proposed stricter rules for AI security, while Trump and House Speaker Mike Johnson have favored voluntary measures by industry.

Cairncross spoke before joining a White House lunch that brought together Trump, Vice President JD Vance, Johnson and leading AI executives to discuss how to weigh safeguards against the push for innovation.

The emphasis on agents reflects a practical concern for operators. AI agents often run with service credentials and broad permissions, and in critical infrastructure they may sit alongside operational technology that was never designed with software agents in mind. Cairncross's checklist of authorization, access and location amounts to asking operators to keep an inventory of agents the way they are expected to track user accounts and connected devices.

The Legacy Problem

Cairncross's point about legacy systems addresses a known weak spot. Many utilities and telecom carriers still run equipment and software past the end of vendor support, and AI tools that can find flaws quickly make those systems easier to attack. Because replacements are expensive and service disruptions are hard to schedule, operators often defer upgrades, which is why he said incentives are needed.

What to Watch

Details of any incentive program for replacing unsupported systems, and whether it takes the form of grants, tax treatment or procurement rules, will show how far the administration goes beyond voluntary measures. Also watch for new public-private arrangements to deploy AI defenses in specific sectors, and for whether congressional proposals for binding AI security requirements advance despite the White House's preference for voluntary commitments.