Britain’s cyber agency laid out plans on Tuesday for what it called “a national scale, sovereign defense capability” that would use agentic AI systems to discover and fix cybersecurity weaknesses across government networks and critical national infrastructure.

The capability, called Cyber Shield, is designed to counter a threat the National Cyber Security Centre (NCSC) said could see attackers “move at machine speed and greater scale, reducing opportunities for detection and response.”

Adversaries aided by AI, the agency said in a blog post, can already compress reconnaissance and vulnerability discovery from weeks into minutes.

“This has the potential to overwhelm traditional defenses and increase the risk of advantage shifting towards the attacker,” the NCSC said. “Developing viable solutions that scale and execute at the pace we need in the modern era is the remit of the Cyber Shield.”

The agency has separately warned of an AI-driven “patch wave” — a surge of newly discovered vulnerabilities emerging faster than most organizations can fix them — and a recent alert from GCHQ said it was likely both offensive and defensive cyber capabilities would be fundamentally transformed within just months.

Cyber Shield was referenced by GCHQ director Anne Keast-Butler in her inaugural annual lecture earlier this year, where she said the agency would “hardwire” agentic AI into machine-speed cyber defense and warned of a narrowing window for the U.K. to stay ahead of its adversaries.

At the heart of the plan is a model of paired “red” and “blue” AI agents — the former probing systems for weaknesses, the latter defending them in real time — operating across critical national infrastructure under the control of the organizations that own them.

The NCSC said Cyber Shield will require six core functions, ranging from automated scanning of British networks — which already exists in some form — to fully autonomous fixing of vulnerabilities, which does not. Some of these functions, the agency acknowledged, “present challenges which will need significant progress in research to unlock.”

The capability is intended to be delivered “in association or partnership with leading frontier AI capabilities, cyber defense organizations and academia.” Initial testing would begin with network defenders across government and critical U.K. sectors before the agency attempts to transition to commercially scalable solutions.

The NCSC set out a “test, iterate, scale” rollout and attached no timeline to the program. It said Cyber Shield cannot be built by the government alone, issuing an open invitation to academia, critical infrastructure operators, frontier AI labs and the cyber defense sector to help develop the blueprint and called on interested parties to get in touch.