The UK’s National Cyber Security Centre says it is highly likely AI-powered malware will be used in a ransomware attack on UK critical national infrastructure between now and 2028. The forecast appears in the agency’s 2026 annual review, published on 3 November. It describes a future risk, not a reported attack, and the NCSC frames the immediate danger as familiar security weaknesses becoming easier to find and exploit at speed. [NCSC Annual Review 2026] [NCSC launch statement]

What the NCSC forecast says

The report’s specific judgment is that it is “highly likely” that, between now and 2028, AI-powered malware will have been used in a ransomware attack on UK CNI. In the same section, the NCSC says it is highly likely that by 2028 China, Russia and other nation states will routinely use AI cyber capabilities to pursue geopolitical goals, with the UK facing increased scale and speed in sophisticated cyber attacks. Those are related but distinct assessments: the ransomware sentence is about AI-powered malware and UK critical infrastructure; the nation-state sentence is broader. [NCSC Annual Review 2026]

The NCSC launch notice summarizes the malware warning as a ransomware attack on Britain’s critical national infrastructure by 2028. Reuters independently reported the same core forecast from the annual report. The Guardian also described the expected use of AI in ransomware against CNI. These reports corroborate what the document says; they do not turn a probability judgment into certainty or establish that an incident has taken place. [NCSC launch statement] [Reuters] [The Guardian]

“Highly likely” is a forecast, not a countdown

The wording should be read as an assessment of likelihood, not a scheduled event. The public-facing review gives no precise probability percentage for this individual forecast, and does not specify a date, named target, attacker, or attack method beyond ransomware involving AI-powered malware. Its time horizon is “between now and 2028.” The NCSC’s public summary says “by 2028”; either way, the source does not claim that a particular service will be struck on a particular day. The distinction matters: preparedness follows from risk, not from treating a forecast as a prediction of one known attack.

What AI changes—and what it does not

The review’s explanation is operational rather than science-fictional. It says AI, automation and computing advances are already helping some threat actors identify vulnerabilities, target victims and work more efficiently. It expects criminal adoption to increase the scale and efficiency of ransomware operations, social engineering and vulnerability exploitation. The report also says open-source AI tools developed to assist security researchers may be repurposed to orchestrate and automate criminal workflows, potentially reducing criminals’ costs and increasing attack volumes against less resilient targets. These are the agency’s threat assessments; the report does not say every tool will be misused or that AI independently creates a new class of attack. [NCSC Annual Review 2026]

At the launch, NCSC chief executive Richard Horne was due to emphasize that AI had not invented new ways to attack. Rather, vulnerabilities tolerated over time may now be discovered and exploited at a scale previously difficult to achieve. That framing keeps the forecast grounded in familiar weaknesses: exposed services, unpatched systems, inadequate investment and poor resilience. AI can affect the speed and economics of finding or using such weaknesses, while the underlying need for secure systems remains. [NCSC launch statement] [Reuters]

Who and what falls within scope

The claim concerns UK critical national infrastructure, not every UK business or every digital service. The NCSC report does not identify which infrastructure sector would be affected. The Guardian notes that the UK’s CNI definition spans energy, transport, defence, food, health and finance; that list supplies useful context, but it is not a list of predicted targets in the NCSC forecast. Critical infrastructure also depends on complex suppliers and connected systems, so disruption could have effects beyond an individual operator. That is a reason to take resilience seriously, not evidence that all sectors face equal risk or that disruption is inevitable. [The Guardian] [NCSC Annual Review 2026]

Nor does the ransomware forecast assign blame. Separate language in the annual review assesses nation-state intentions and AI use, including an assessment that China is highly likely to have the intent and be best positioned to use AI against CNI and other strategic targets. That statement is a separate NCSC judgment about state capability and intent; it is not an attribution of the forecasted ransomware attack to China, Russia, or any government. Reuters reported the China assessment, while noting that Beijing, Moscow and Tehran have previously denied accusations of malicious cyber activity. [NCSC Annual Review 2026] [Reuters]

The agency’s practical recommendations

The NCSC’s answer is to address foundational cyber security and resilience, not wait for a bespoke AI defence. Its launch statement says organisations should invest adequately in technology, avoid connecting systems to the internet unnecessarily, and use Cyber Essentials as a starting point rather than the finish line. It calls for resilience proportionate to an organisation’s size, complexity and risk, and urges organisations to make use of existing tools, standards and frameworks that help defend against conventional as well as AI-assisted attacks. [NCSC launch statement]

The annual review’s board-level messages reinforce those basics: maintain good cyber security fundamentals, involve senior leaders in decisions, and understand how agentic AI tools are used and what access they have to systems and data. It also warns that unmanaged “shadow AI”—employees using unsanctioned tools and transferring sensitive information—can reduce an organisation’s visibility into how information is handled. The risk is not limited to attackers using AI; careless adoption inside an organisation can create its own exposure. [NCSC Annual Review 2026]

Prepare for disruption, not only prevention

Prevention cannot promise that every intrusion will be stopped. The NCSC highlights recovery guidance for organisations whose critical digital systems have significant operational consequences if lost. Its three phases cover immediate containment and assessment, setting up a recovery programme over the first days or weeks, and rebuilding operations. The inclusion of recovery planning is important because a ransomware incident can become an operational crisis even when an organisation has security controls. Business continuity, decision authority, communications and restoration priorities should be considered alongside vulnerability management. The report’s direction is preparedness and resilience; it does not prescribe a single technical product or universal timetable for every organisation. [NCSC Annual Review 2026]

What the warning means for readers

For the public, the report is a signal about the importance of reliable essential services, not a reason to infer that a cyberattack is imminent at a named hospital, airport or utility. For organisations operating or supporting CNI, the actionable issue is whether basic controls, exposure management and recovery arrangements can keep pace with a faster threat environment. The NCSC points to investments in national cyber defence, including a Cyber Shield, a National Exposure and Real-Time Vulnerability Centre and an AI Cyber Security Centre of Excellence. These government programmes complement, but do not replace, individual organisations’ responsibility to secure their systems. [NCSC launch statement]

The forecast is newsworthy because it combines a firm likelihood assessment with a defined window and a high-impact target class. But its limits should travel with the headline: the public report supplies no numerical probability, no specific victim, and no attribution for a prospective AI-enabled ransomware incident. Independent reporting confirms the statement, while the primary review remains the source for the exact formulation and its surrounding recommendations. The most defensible reading is both serious and bounded: the agency sees such an attack as highly likely by 2028, and urges organisations to strengthen fundamentals now.

Questions and answers

What exactly does the NCSC say is highly likely?

It assesses that AI-powered malware will have been used in a ransomware attack on UK critical national infrastructure between now and 2028. The statement is a forecast, not a report that the event has already occurred. [NCSC Annual Review 2026]

Does the forecast name a target or attacker?

No. The ransomware sentence names neither a particular infrastructure operator nor an attacker. The review discusses state actors elsewhere, but those assessments should not be mistaken for attribution of a future ransomware event.

Does the report say AI creates entirely new attack methods?

No. At the review launch, NCSC chief executive Richard Horne’s stated argument was that AI could help discover and exploit weaknesses that organisations have tolerated, increasing the scale and speed of activity. [NCSC launch statement]

What should organisations do first?

The NCSC urges adequate technology investment, avoiding unnecessary internet exposure, and treating Cyber Essentials as a baseline rather than a complete security programme. It also stresses proportionate resilience, good fundamentals, senior oversight and plans to respond and recover. [NCSC launch statement] [NCSC Annual Review 2026]

Is the forecast a guarantee that an attack will happen in 2028?

No. “Highly likely” expresses the NCSC’s assessment, while the report does not provide a percentage, precise date or named incident. It is a warning to prepare for risk, not a guarantee or a countdown.

References