The United States is moving to embed artificial intelligence across critical infrastructure networks in a bid to strengthen cyber defenses, National Cyber Director Sean Cairncross told telecom and broadband leaders on Tuesday.[1][2][15] Speaking at an event hosted by industry group USTelecom, Cairncross said his office is working closely with operators to deploy frontier AI models “throughout sectors as quickly as possible” so they can be used to secure systems rather than undermine them.[1][2]
Cairncross warned that as AI agents are woven into operational technology and enterprise environments, CEOs need clear visibility into which agents are running, who authorized them, what data and systems they can access, and how they interact with complex supply chains.[1][2][15] He framed control over machine-driven agents as a board-level risk issue, arguing that misconfigured or opaque automation could become a new class of insider threat or backdoor if enterprises fail to map and audit these systems.[1][5]
The remarks came just before a White House lunch with top AI executives, President Donald Trump, Vice President J.D. Vance and House Speaker Mike Johnson, where officials were expected to debate whether strict regulatory guardrails are needed or whether voluntary commitments from industry are sufficient.[1][4][6] Cairncross aligned himself with the administration’s preference for fast deployment over sweeping rules, saying the goal is to secure advanced AI “without imposing broad regulations that could slow technology development” or create a top-down regime applied uniformly across sectors.[4][10]
Cairncross pointed to what he called the “Mythos moment” in the spring—when the Mythos AI model debuted with highly advanced coding and vulnerability-discovery capabilities—as crystallizing the urgency of the government’s work to shape AI policy and security pathways.[1][6][15] Since Mythos’ release, officials have convened repeated meetings with tech companies and security experts, trying to manage rollout of models whose hacking proficiency rivals that of top human specialists while avoiding overregulation that might push innovation offshore.[6][3]
The national cyber director has repeatedly argued that AI is exposing long-standing weaknesses in critical systems rather than creating entirely new categories of risk, noting that automated tools are surfacing latent software flaws and configuration errors that defenders failed to address for decades.[5][14] At forums including Black Hat USA 2026, he has described AI-enabled red-teaming and blue-teaming at “machine speed” as both a powerful defensive capability and a way for adversaries to lower the barrier to building sophisticated exploits against U.S. infrastructure.[3][9]
Despite the high-profile focus on frontier models, Cairncross said the less glamorous work of replacing aging, unsupported systems in utilities, telecom and other critical sectors remains central to any AI-driven security strategy.[1][4] He has floated the idea of new incentives to accelerate modernization, arguing that embedding AI into brittle legacy environments will only magnify existing vulnerabilities unless organizations invest in patching, asset management and secure-by-design architectures that can support advanced automation.[7][12][13]
For now, the message from the White House’s cyber office is that AI will be integrated deeply into the nation’s vital systems, but that success hinges on executives treating AI agents, model pipelines and aging infrastructure as intertwined risks rather than separate technology projects.[1][5][13] CEOs who rush to deploy powerful models without parallel investments in visibility, governance and modernization, Cairncross suggested, may find that the same tools meant to protect critical services can be turned against them.[1][3][6]
References
- US is looking to weave AI into critical infrastructure for …
- US is looking to weave AI into critical infrastructure … – ZeroHour
- NightVision with Sean Cairncross — National Cyber …
- Cairncross: White House AI Talks Aim for Secure, Rapid …
- Governments ‘buying time’ in race between innovation … – CyberScoop
- National Cyber Director Sean Cairncross is leading the …
- National cyber director expands on Trump administration’s vision for …
- Cyber Director Sean Cairncross on AI and Cybersecurity at Black …
- National cyber director lays out White House plans to secure AI …
- White House cyber shop is crafting AI security policy …
- White House walks tightrope on securing AI without stifling …
- Sean Cairncross Archives | CyberScoop
- Mythos (AI model): news timeline & CVEs – ZeroHour
