Summary

• The NCSC assesses that AI-powered malware is highly likely to be used in a ransomware attack against UK critical national infrastructure by 2028.
• Between September 2025 and August 2026, the agency received 1,831 cyber incident reports and supported responses to 337 incidents, including 80 classified as nationally significant.
• The assessment identifies known vulnerabilities, legacy technology, insecure internet connections, and supply chain weaknesses as existing routes that AI could help attackers exploit more efficiently.

Britain’s National Cyber Security Centre has warned that artificial intelligence will accelerate cyberattacks against UK organisations, assessing that AI-powered malware is highly likely to be used in a ransomware attack on critical national infrastructure by 2028.

The agency’s State of the Cyber Nation review, published on 3 November, identifies a growing threat from attackers using AI to discover vulnerabilities, identify potential targets, and conduct established forms of intrusion more efficiently. Its assessment suggests that many organisations remain exposed through weaknesses that could already be addressed using existing security practices.

The NCSC expects Russia, China, and other nation states to use AI routinely in cyber operations by 2028, while criminal groups are likely to exploit similar technology to increase the scale and efficiency of ransomware, social engineering, and vulnerability exploitation.

The warning accompanies new figures covering the agency’s reporting year from 1 September 2025 to 31 August 2026. During that period, the NCSC received 1,831 reports of cyberattacks affecting UK organisations, with its incident management team providing support in 337 cases. Of those, 80 were classified as nationally significant.

Existing vulnerabilities face faster exploitation

The NCSC’s analysis distinguishes the growing use of AI from the development of entirely new methods of intrusion. Many successful attacks continue to begin with known software weaknesses, stolen credentials, poorly secured systems, and services exposed unnecessarily to the internet.

AI can assist attackers by reducing the effort required to identify suitable targets and examine potential weaknesses. Tools capable of analysing large quantities of technical information may allow individuals or groups to assess more systems than would be practical using exclusively manual methods.

Attackers may also use AI to prepare convincing phishing messages, interpret information collected during reconnaissance, or automate parts of existing attack workflows. These capabilities can increase operational efficiency without removing the need for experienced individuals to direct and adapt the activity.

The agency’s assessment indicates that human operators are highly likely to remain important in cyber operations. Although AI can assist with particular tasks, its contribution varies across the stages of an attack, with skilled attackers often better placed to exploit its capabilities.

That distinction affects the response required from organisations. Weaknesses in software maintenance, authentication, and network access remain relevant even where attackers use more advanced tools to discover and exploit them.

Technology estates containing unsupported equipment or poorly documented systems may be especially difficult to defend. Organisations can struggle to identify which software versions are installed, which services are accessible externally, and whether a newly discovered vulnerability affects equipment supporting important operations.

Legacy infrastructure creates persistent exposure

Critical national infrastructure includes services whose disruption could affect public safety, economic activity, or essential functions. Operators in sectors such as energy, water, telecommunications, and transport depend on combinations of conventional business technology, industrial control systems, specialist equipment, and external suppliers.

Some operational systems have long service lives and were installed before remote connectivity became common. Equipment that was originally isolated may subsequently have been connected to management platforms or external maintenance services, introducing access routes that require protection.

Updating such systems can involve operational difficulties because an interruption may affect production, distribution, or service availability. Maintenance therefore requires coordination between technical teams, equipment manufacturers, and managers responsible for keeping services running.

The NCSC considers it likely that attackers will apply AI capabilities against known vulnerabilities in legacy critical infrastructure technology by 2028. It also identifies additional exposure associated with the adoption of AI tools within operational environments, where new software and connections may increase the number of systems requiring protection.

Supplier relationships create another route for compromise. A contractor responsible for maintaining equipment may hold privileged access to several customer environments, while software supplied by an external organisation can introduce vulnerabilities outside the immediate control of its users.

Security teams must understand those dependencies and establish how access is granted, monitored, and withdrawn. Improving the protection of directly owned systems may be insufficient where contractors or technology providers retain unnecessary permissions.

Incident figures reveal continuing pressure

The NCSC’s annual review records 337 incidents requiring support from its incident management team. That total includes 80 nationally significant incidents, reflecting events that required coordination because of their importance to the UK’s security or wider interests.

The agency also issued Early Warning notifications to approximately 8,500 organisations after identifying activity suggesting a possible compromise. These notifications are intended to help recipients investigate suspicious events before they develop into more serious incidents.

The figures cover different forms of activity and should not be combined into a single count of confirmed successful attacks. Reports received by the NCSC, incidents requiring its intervention, and preventive notifications measure different stages of detection and response.

Annual totals also vary with reporting behaviour, the seriousness of individual incidents, and the effectiveness of early intervention. The NCSC cautions against treating changes in reported incidents as a straightforward measure of whether the overall threat is increasing or decreasing.

Its incident investigations nevertheless reveal recurring patterns. Attackers continue to exploit known vulnerabilities, abuse credentials, and take advantage of inadequately secured systems, often obtaining initial access through methods that organisations could have made more difficult.

Ransomware remains particularly disruptive because recovery can involve more than restoring encrypted files. Organisations may need to investigate how attackers entered their systems, determine whether sensitive information was stolen, rebuild compromised infrastructure, and confirm that restored services can be trusted.

Where business applications or operational technology support physical services, interruptions can affect production and service delivery even if the attackers initially intended to obtain money rather than cause physical damage.

AI scenarios extend to 2028

The NCSC has developed three scenarios describing possible developments in AI-enabled cyber threats through 2028. Its baseline scenario anticipates an increase in the global threat as attackers gain access to more capable tools, while a pessimistic scenario examines the consequences of offensive capabilities developing faster than defensive measures.

An optimistic scenario considers the possibility that defensive improvements and limits on offensive AI development allow organisations to establish more effective countermeasures. The scenarios are designed to support planning rather than predict specific future incidents.

Within its central assessment, the NCSC judges it highly likely that AI-powered malware will have been used in a ransomware attack on UK critical national infrastructure by 2028. This is a forecast about the expected use of AI within an attack, not a statement that such an incident has already occurred or that an autonomous system will inevitably disable essential services.

The agency also considers it almost certain that many systems will remain vulnerable to tools capable of identifying and exploiting existing weaknesses. As automated targeting becomes cheaper, attackers may be able to replace lost access more readily and pursue larger numbers of potential victims.

AI may simultaneously improve defensive activity by assisting with vulnerability analysis, threat detection, and the assessment of suspicious behaviour. Those capabilities depend on reliable information and suitable operational procedures, while automated security recommendations still need to be assessed before changes are made to important systems.

Resilience demands operational investment

Dr Richard Horne, chief executive of the NCSC, has warned that organisations can no longer rely on prolonged underinvestment in security and technology without facing greater consequences. His remarks accompanying the report call for stronger management of internet-connected systems, better maintenance, and investment in resilience before incidents occur.

The agency identifies Cyber Essentials as a starting point for organisations seeking to reduce common risks, while making clear that businesses with more complicated operations or greater exposure require additional safeguards.

Improving security can involve substantial operational work. Replacing unsupported equipment, restricting unnecessary remote access, maintaining software, and reviewing supplier permissions require technical expertise and decisions about investment, maintenance schedules, and service continuity.

Organisations also need to understand which systems support their most important functions. An accurate asset inventory helps teams identify vulnerable equipment, determine who is responsible for updates, and establish whether changes have been completed.

Recovery arrangements provide protection when preventive controls fail. Tested backups, incident response procedures, and plans for maintaining essential operations can reduce the duration and consequences of a compromise, particularly where external suppliers form part of the affected service.

The NCSC’s assessment follows separate government plans for investment in AI-assisted cyber defence, including the National Cyber Shield and a vulnerability assessment centre. Those programmes address national capabilities, while the new report concentrates on how the threat is developing and the existing weaknesses organisations must address.

With AI expected to increase the speed and volume of cyber operations, the NCSC is directing attention towards the maintenance and protection of systems already in use. Its annual review shows that attackers continue to exploit familiar weaknesses, while the new assessment indicates that the same weaknesses could be targeted more efficiently over the next two years.