The UK’s National Cyber Security Centre (NCSC) says attackers will highly likely use AI-powered malware in a ransomware attack on critical national infrastructure by 2028. The warning is in the NCSC annual review, The State of the Cyber Nation, published on 3 November.
The NCSC, part of the GCHQ spy agency, also expects hostile states to lean on AI.
“By 2028, it is highly likely China, Russia, and other nation states will routinely use AI cyber to achieve their geopolitical goals,” the NCSC wrote.
The review calls China’s activity the most sophisticated and sustained cyber challenge facing the UK. It names Russia as the most audacious cyber threat today. Iran-linked groups, it adds, show how cyber operations can cause indiscriminate disruption at scale.
One case study covers Poland. The UK and EU blamed Russia’s FSB Centre 16 for an attack on Poland’s energy grid in December 2025. Had it worked, 500,000 people could have lost power, the NCSC said.
The NCSC also named China as the state best placed to use AI against critical infrastructure. It said China highly likely has the intent to do so.
A year of incidents
The review covers 1 September 2025 to 31 August 2026. In that time, the NCSC’s incident team received 1,831 reports of cyber attacks. Of those, 337 needed its support, and the NCSC rated 80 as nationally significant. Its Early Warning service also alerted 8,500 organisations to signs of a suspected compromise.
Most successful attacks still used familiar methods: known flaws in internet-facing systems, stolen credentials and social engineering. For large organisations, 88% of incidents started in one of four ways. Those were exploiting public-facing applications, phishing, misuse of valid accounts and exploiting remote services.
The NCSC cites government research that puts the cost of malicious cyber activity at £14.7bn a year for UK organisations. Jaguar Land Rover reported about £196mn in direct costs from its 2025 attack. The NCSC said that attack contributed to UK growth of only 0.1% in the third quarter of 2025.
The NCSC said it is “almost certain” that AI will act as a force multiplier for skilled attackers. AI will mostly speed up and scale existing techniques rather than create new ones. Its incident team has begun to see attackers add AI tools, but none of those uses enabled a new class of attack.
The agency expects a “patch wave” as AI uncovers old flaws in software. It urged organisations to plan for a rush of security updates, some of them critical.
The review also sets out a pessimistic scenario for 2028, which the NCSC said is not a prediction. In it, an automated attacker built on frontier AI slips out of its operator’s control. The NCSC called it a “realistic possibility” that such an entity could copy itself and attack at will.
The NCSC still believes AI will ultimately be a “net positive” for cyber security. Getting there requires urgent action now, it said.
“Tech on the cheap”
NCSC chief executive Richard Horne launched the review with a speech in London.
“In the coming months, we face a real prospect of an unprecedented rise in the number of cyber incidents. But we must be clear, that is not because AI has invented new ways to attack us,” Horne said.
He blamed old flaws that attackers can now find and exploit at a new scale and pace.
“For decades we have collectively got away with tech on the cheap,” Horne said.
Horne said UK military networks face attempts to break in every day. He also cited the disruption of refrigeration units across Europe, including the supermarket freezer attacks.
The NCSC issued more than 110,000 notifications and alerts to UK organisations after the PHP2Shell vulnerability, Horne said.
The review describes Project DROMOS, which tested frontier AI models on government code. After a four-week pilot, it moved into selected government systems. There it uncovered hundreds of potential security issues, including critical vulnerabilities.
The NCSC and the culture and digital department plan to build the national Cyber Shield over five years. The government funded it in last week’s AI cyber defence budget. Horne urged organisations to treat Cyber Essentials, the government’s basic security certification, as a starting point rather than the goal.
“We’re using the power of AI to help stay ahead of those who seek to do us harm,” Security Minister Dan Jarvis said in a statement.