The NCSC AI malware warning is the starkest assessment yet from Britain's cyber defenders about how artificial intelligence will change the threat facing the country's most important services. In its annual report, the National Cyber Security Centre (NCSC) — part of GCHQ — said AI-powered malware is highly likely to be used in a ransomware attack on Britain's critical national infrastructure by 2028, and warned that the coming months could bring an unprecedented rise in the number of cyber incidents. The warning matters because critical national infrastructure covers the systems everyday life depends on: energy, transport, defence, food, health and finance. The NCSC named power stations, airports and hospitals as the kind of facilities in scope, which means the warning is not about abstract technology risk but about the electricity, travel and healthcare services millions of people rely on every day.
The report, called State of the Cyber Nation, was published on Tuesday 3 November 2026 and was launched by NCSC chief executive Dr Richard Horne. Launching the report, Dr Horne warned of "a real prospect of an unprecedented rise in the number of cyber incidents" in the coming months. He also said that the era of "tech on the cheap" at the expense of security is over, a direct challenge to organisations that have treated cyber security as an optional extra rather than a core cost of doing business. According to the NCSC, artificial intelligence will expose vulnerabilities that have been tolerated for far too long, forcing organisations to fix weaknesses they have previously been able to ignore. The message is that AI is not only a tool for attackers; it is also a spotlight that reveals how fragile many existing systems are.
What the NCSC AI Malware Warning Says About Ransomware by 2028
The centrepiece of the NCSC AI malware warning is its assessment of ransomware. The NCSC said it is highly likely that AI-powered malware will be used in a ransomware attack on Britain's critical national infrastructure by 2028. Ransomware is a form of malicious software that locks a victim's systems or data until a ransom is paid, and attacks on hospitals, power companies and transport operators can cause disruption far beyond the organisation that is hit. The NCSC's assessment is that AI will make such attacks easier to launch, harder to detect and more effective once they begin. The report also said hostile states and criminal gangs are highly likely to use AI to increase the scale and efficiency of their operations by 2028, including ransomware, social engineering and the exploitation of software vulnerabilities. Social engineering means tricking people into handing over access or information, and AI can make those tricks far more convincing, for example by generating realistic messages in flawless language or by impersonating a trusted colleague.
The NCSC singled out two countries in particular. It said Russia and China will highly likely and routinely use AI in cyber operations by 2028. That assessment reflects a broader view inside Western intelligence agencies that AI is becoming a standard tool of statecraft, not an experimental novelty. The report also set out a pessimistic scenario for 2028 in which there is a significant increase in the frequency and sophistication of cyberattacks worldwide, with AI making it harder to attribute attacks to their authors. Attribution matters because governments and companies need to know who attacked them before they can respond, deter future attacks or seek help from allies. If AI makes attribution harder, defenders lose one of their most important tools. The report also raised a realistic possibility that by 2028 there could be an autonomous attacker — a system that could self-replicate and conduct cyberattacks at will, without a human operator directing each step.
Why the NCSC AI Malware Warning Focuses on Critical Infrastructure
Critical national infrastructure is the term the UK government uses for the assets and systems that keep the country running. The NCSC AI malware warning applies to energy, transport, defence, food, health and finance, and the report specifically named power stations, airports and hospitals as examples of the facilities in scope. An attack on any of these would not simply be a data breach; it could interrupt electricity supplies, ground flights or delay medical treatment. That is why the NCSC treats the threat to this sector differently from attacks on ordinary businesses. A ransomware attack on a hospital, for instance, can force staff back to paper records, delay operations and put patients at risk, while an attack on a power station or an airport can cascade into wider disruption across the economy. The NCSC's point is that AI lowers the barrier to carrying out such attacks, so the number of actors capable of attempting them is likely to grow.
The report covers the period from 1 September 2025 to 31 August 2026, the NCSC's tenth year of operation, and the numbers it contains show how busy the centre already is. During that period, the NCSC's incident management team received 1,831 reports of cyber attacks, supported 337 incidents and classed 80 of them as nationally significant. Those figures are the baseline against which the warning should be read: even before AI-powered malware becomes common, the UK is already dealing with hundreds of serious incidents a year, and dozens that are significant enough to affect the nation as a whole. The NCSC's concern is that AI will multiply both the volume and the sophistication of attacks, stretching defenders who are already working at a high tempo. For readers who want to understand how AI is already reshaping cyber defence, our explainer on Google's AI model for cyber defence looks at the defensive side of the same technology.
What Cyber Shield Means for the NCSC AI Malware Warning
Alongside the report, the UK government announced a package of new funding aimed at strengthening the country's cyber defences. The centrepiece is £60 million for the first phase of a national Cyber Shield, a programme intended to protect critical national infrastructure. The government also announced £70 million for a new National Exposure and Real-Time Vulnerability Centre, £40 million for a National Agentic Security Unit, and £24 million to strengthen public-sector cyber defences using AI. A new AI Cyber Security Centre of Excellence is also being set up in Manchester, run by the NCSC. The funding totals £194 million across the four named items, and the Manchester centre signals that the government wants cyber expertise spread beyond London and the south-east. The National Exposure and Real-Time Vulnerability Centre is intended to give the country a clearer picture of which systems are exposed to attack at any given moment, while the National Agentic Security Unit appears aimed at the risks posed by autonomous AI agents — the same kind of self-directed systems the NCSC warned could one day attack without human direction.
The funding announcement is significant because it shows the government is treating the NCSC AI malware warning as a call to action rather than a prediction to watch from the sidelines. Cyber Shield is explicitly aimed at the critical infrastructure the report identifies as most at risk, and the vulnerability centre is designed to close the gap the NCSC identified between known weaknesses and fixes that have been delayed. The £24 million for public-sector defences using AI reflects a growing belief that defenders must use the same technology as attackers: if criminals and hostile states are using AI to find vulnerabilities faster, the organisations defending hospitals, councils and government departments need AI tools of their own. Readers following how the UK government is spending on public services can also read our breakdown of the 2026 UK budget and its seven changes for households.
How AI Changes Ransomware, Social Engineering and Vulnerability Exploitation
To understand the NCSC AI malware warning, it helps to understand the three attack methods the report highlights. Ransomware is the most familiar: software that encrypts a victim's files and demands payment for the key. AI can make ransomware more adaptive, helping it find the most valuable data, spread faster through a network and evade detection tools that rely on recognising known patterns. Social engineering is the human side of cyberattacks — persuading someone to click a link, share a password or approve a payment. AI can generate convincing messages at scale, in any language, tailored to an individual target using information gathered from public sources. Vulnerability exploitation means taking advantage of flaws in software that have not yet been patched. AI can scan vast amounts of code far faster than human researchers, finding weaknesses that would otherwise have gone unnoticed — which is exactly why the NCSC said AI will expose vulnerabilities tolerated for far too long. The same capability can help defenders, but the NCSC's assessment is that attackers will benefit first and fastest.
One of the report's most striking ideas is the autonomous attacker: a system that could self-replicate and conduct cyberattacks at will. This is not a claim that such a system exists today, but a warning that by 2028 it is a realistic possibility. An autonomous attacker would not need a human to choose targets, write malware or decide when to strike, which would make attacks faster, cheaper and harder to predict. The NCSC also warned that AI will make attribution harder, meaning it will be more difficult to prove who was behind an attack. That matters because attribution underpins everything from law enforcement to international diplomacy: without it, governments struggle to impose consequences on attackers, and companies struggle to know whether they are facing a criminal gang, a hostile state or an opportunist. The combination of more attacks, faster attacks and harder attribution is what makes the 2028 scenario the report describes so concerning.
What Organisations and Individuals Should Do After the NCSC AI Malware Warning
The NCSC AI malware warning is aimed primarily at organisations, especially those running critical infrastructure, but its lessons apply more widely. The report's underlying message is that security can no longer be treated as an afterthought or a cost to be minimised. Dr Horne's remark that the era of "tech on the cheap" is over is a warning to boards and senior managers that underinvestment in cyber security now carries a growing risk. For organisations, the practical implications include keeping software patched promptly, because AI will find unpatched vulnerabilities faster; training staff to recognise AI-generated social engineering, which is harder to spot than the clumsy phishing emails of the past; and having tested plans for recovering from a ransomware attack, including backups that are kept separate from the main network so they cannot be encrypted too.
For individuals, the advice is more familiar but no less important: use strong, unique passwords, turn on multi-factor authentication where it is offered, and be sceptical of unexpected messages that ask for money, passwords or urgent action — even if they appear to come from a trusted source. The NCSC's figures show that the threat is already real: 1,831 reports of cyber attacks in a single year, 337 incidents supported and 80 classed as nationally significant. The warning is that AI will make those numbers worse unless defences improve at the same pace. The Cyber Shield funding, the Manchester centre and the new vulnerability centre are the government's answer, but the NCSC's report makes clear that technology alone is not enough — the organisations that run Britain's power stations, airports and hospitals need to treat cyber security as a core part of their job, not an optional extra. Cyber attacks on everyday services are not a distant possibility; as shown by the recent cyberattack on supermarket freezers across seven European countries, disruption to ordinary life is already happening. The NCSC AI malware warning is, in the end, a call to act before 2028 arrives rather than after it.