UK Cyber Agency Expects AI-Powered Ransomware Attack on Critical Infrastructure by 2028

The UK's National Cyber Security Centre has warned that Britain's critical national infrastructure should prepare for a ransomware attack involving AI-powered malware by 2028.

The assessment, published on Tuesday in the agency's annual State of the Cyber Nation report, judges it highly likely that hostile actors will use artificial intelligence over the next two years to make their operations larger and more efficient.

What the NCSC is warning about

The NCSC, which sits within GCHQ, the UK's signals intelligence agency, identified several areas where it expects AI to give attackers an edge. These include ransomware campaigns, social engineering aimed at tricking staff into handing over access, and the exploitation of flaws in software.

The report frames the coming period as a point of decision for the country. The agency's message is that the outcome for national cyber resilience has not yet been determined, and that choices made by organisations and government now will shape how damaging the next wave of attacks proves to be.

Richard Horne, the NCSC's chief executive, used prepared remarks for the report's launch to stress that the danger is less about novel techniques than about familiar shortcomings. In his account, AI has not created fresh methods of attack. Instead, it allows weaknesses that have been left unaddressed for years to be found and exploited far faster and on a far larger scale than was previously possible.

Horne said the UK faced the real prospect of an unprecedented increase in the number of cyber incidents in the coming months. He was blunt about the root cause, arguing that for many years organisations across the country had escaped the consequences of underspending on technology.

A call for basic discipline

Horne's prescription was notably unglamorous. He urged organisations to invest properly in their technology, to stop connecting systems to the internet when there is no need to do so, and to treat Cyber Essentials, the government-backed certification scheme, as a minimum baseline rather than a finishing line.

That emphasis reflects a long-standing frustration among security professionals. Many successful breaches exploit unpatched software, exposed services or weak authentication rather than sophisticated zero-day techniques. If AI tools make it cheap to scan for and weaponise those gaps at scale, the cost of neglecting routine maintenance rises sharply.

Horne also sought to place responsibility at board level. He warned that leaders who fail to act could find themselves explaining to investors, employees and customers why shops have run out of food, power has gone off or businesses have collapsed. The framing ties cyber security directly to physical outcomes, which is precisely the concern with attacks on critical infrastructure such as energy, water, transport and food supply chains.

Government funding and new institutions

The report follows a set of funding announcements made by the government the previous week. These included £60m (about $80m) for the first phase of a national Cyber Shield and £70m ($93m) for a new National Exposure and Real-Time Vulnerability Centre, intended to identify weaknesses across critical infrastructure.

The government also announced a new AI Cyber Security Centre of Excellence in Manchester, which will be run by the NCSC.

Taken together, the measures suggest officials are focusing on visibility: understanding where vulnerabilities sit across the national infrastructure before attackers find them. A real-time exposure centre addresses the same problem Horne described, namely that defenders often lack a clear picture of what is connected, what is outdated and what is exposed.

Whether roughly £130m is sufficient for that task is a separate question. Critical infrastructure in the UK is spread across public bodies and private operators, many running legacy systems that are expensive and disruptive to replace. Central funding can improve monitoring and coordination, but much of the remediation work will fall on individual organisations and their budgets.

Evidence that attackers are already using AI

The NCSC's warning comes amid mounting signs that criminals are putting AI to practical use. In May, Google's Threat Intelligence Group reported that cybercriminals had used an AI model to discover and weaponise a previously unknown flaw in a popular open-source administration tool. The vulnerability allowed attackers to bypass two-factor authentication. According to Google, this marked its first verified case of attackers using AI to build a zero-day exploit in real-world conditions.

That case illustrates the dynamic Horne described. Finding exploitable bugs in widely used software has traditionally required significant skill and time. If AI models can compress that effort, the pool of actors capable of producing working exploits could widen considerably.

Yet the picture is not uniform. A Cambridge University study published the same month examined more than 90,000 threads on cybercrime forums and concluded that the majority of offenders were applying AI to phishing and spam, not to sophisticated intrusions. That suggests the bulk of current AI use in the criminal ecosystem is about volume and polish, such as more convincing scam emails, rather than breakthroughs in offensive capability.

The two findings are not necessarily in tension. A small number of capable groups could use AI for high-end exploitation while the wider criminal market uses it to scale up cruder fraud. Both trends would increase pressure on defenders, though in different ways.

Risks and counterpoints

Forecasts of this kind carry inherent uncertainty. The NCSC's language reflects a probabilistic judgement rather than a prediction of a specific event, and the timeline to 2028 leaves room for both attackers and defenders to adapt. Defensive teams also have access to AI tools that can help with vulnerability discovery, log analysis and patch prioritisation.

There is also a risk that headline warnings about AI distract from the more mundane fixes Horne himself emphasised. If organisations interpret the threat as a futuristic problem requiring exotic solutions, they may underinvest in the basics, such as patching, network segmentation and limiting internet exposure, that the NCSC regards as the real priority.

UK officials have raised alarms about AI before. In July, Foreign Secretary Yvette Cooper argued that governments must agree on safeguards for AI before frontier systems reshape warfare, crime and society, warning against waiting for a catastrophic event to prompt action. The NCSC report places that broader concern in a narrower, more operational context.

What to watch next

Several developments will indicate whether the UK is closing the gap the NCSC has identified:

• How quickly the National Exposure and Real-Time Vulnerability Centre becomes operational and what it reveals about weaknesses in critical infrastructure.
• Whether further phases of the Cyber Shield receive funding beyond the initial £60m.
• Any rise in reported incidents in the coming months, which would test Horne's warning of an unprecedented increase.
• Further confirmed cases of AI-assisted exploit development from threat intelligence teams.
• Whether boards and regulators begin treating Cyber Essentials as a floor rather than a ceiling.

For now, the agency's central message is less about the novelty of AI than about accountability. The tools available to attackers are improving, but the vulnerabilities they will exploit are, for the most part, ones organisations already know about.