AI malware in the UK in 2026 show a fast-rising threat that official figures do not yet count separately: the NCSC handled 204 nationally significant cyber incidents in the year to August 2025, up from 89, and said hostile states were linked to about 75% of more than 200 attacks on critical national infrastructure in the year to May 2026. Government testing finds frontier AI cyber capability now doubling every 4 months, Google has documented AI-enabled malware families used in live attacks, and 43% of UK businesses (about 612,000) reported a breach or attack, though only 24% of those adopting AI have security practices for it.
AI Malware in UK 2026 – Introduction
AI malware statistics in the UK in 2026 come from three kinds of evidence: official incident counts from the National Cyber Security Centre (NCSC), the government’s annual Cyber Security Breaches Survey, and threat research on AI-enabled malware that is mostly global. No UK agency publishes a count of attacks that used AI malware, so this report separates what is measured, such as incidents and breach rates, from what is assessed, such as how AI is expected to change the threat.
The NCSC says AI will almost certainly make cyber attacks more frequent and more effective through 2027. Its incident figures show a UK already under heavy pressure from ransomware, phishing and state-linked intrusions, including attacks on critical national infrastructure (CNI). Figures run through November 3, 2026, and assessments are labeled as assessments, not counts.
Interesting Facts About AI Malware and Cyber Attacks in the UK in 2026
NCSC NATIONALLY SIGNIFICANT INCIDENTS, YEAR TO AUGUST (1 block = 5 incidents)
2022-23 | ████████████▌ 62
2023-24 | █████████████████▌ 89
2024-25 | ████████████████████████████████████████▌ 204
| Fact | Confirmed Data |
|---|---|
| Nationally significant incidents, Sept 2024 to Aug 2025 | 204, up from 89 (+129%) |
| Highly significant incidents, same period | 18, up from 12 and 4 two years earlier |
| UK businesses reporting a breach or attack, 2025/26 | 43%, about 612,000 |
| Large UK businesses reporting a breach or attack | 69% |
| UK businesses hit by ransomware, 2025/26 | 1%, down from 3% |
| CNI incidents handled by NCSC, year to May 2026 | More than 200, about 75% linked to hostile states |
| Cost of the Jaguar Land Rover attack (CMC estimate) | About £1.9 billion |
| Frontier AI cyber capability doubling time (AISI) | 4 months, down from 8 months |
| UK businesses using or adopting AI with AI security practices | 24% |
| AI-enabled malware families documented by Google in 2025 | 5 |
Source: NCSC Annual Review 2025; DSIT and Home Office, Cyber Security Breaches Survey 2025/2026; NCSC CEO speech, June 2026; Cyber Monitoring Centre; UK Government open letter, April 15, 2026; Google Threat Intelligence Group.
The NCSC says it now handles an average of about four nationally significant incidents a week. Incidents of this kind accounted for 48% of everything the agency managed in 2024-25, compared with about 21% the year before. Total incidents handled stayed roughly flat at 429, so the change is in severity, not volume.
AI features in government warnings as an accelerant, not yet as a counted cause. The April 2026 open letter from the technology and security ministers says a new generation of AI models can find software weaknesses and write exploit code at speed and scale. The most severe UK incidents of 2025, such as the JLR and M&S attacks, were ransomware operations, and the sources reviewed do not describe them as AI-malware attacks.
NCSC Cyber Incident Statistics in the UK: Nationally Significant and Highly Significant Attacks
NCSC INCIDENTS HANDLED PER YEAR (SEPT TO AUG) (1 block = 20 incidents)
2022-23 | ██████████████████▌ 371
2023-24 | █████████████████████▌ 430
2024-25 | █████████████████████▌ 429
| Year (Sept to Aug) | Tips Received | Incidents Handled | Nationally Significant | Highly Significant | Significant Share |
|---|---|---|---|---|---|
| 2022-23 | 2,005 | 371 | 62 | 4 | 16.7% |
| 2023-24 | 1,957 | 430 | 89 | 12 | 20.7% |
| 2024-25 | 1,727 | 429 | 204 | 18 | 47.6% |
Source: NCSC Annual Review 2025. Significant share calculated from published counts. “Nationally significant” covers the top three categories of the UK cyber incident model.
Highly significant incidents have risen for the third year in a row, from 4 to 12 to 18, about 4.5 times the 2022-23 level. These are incidents with the potential to seriously affect government, essential services or the economy, and the NCSC says they often need a cross-government response. The UK has not recorded a Category 1 national cyber emergency.
The NCSC cautions that its data understate the real number of incidents, because many organisations do not report to it. Its review also notes ransomware as the most disruptive risk, with attacks reaching logistics, manufacturing and energy distribution. The agency did not list which incidents fell in the highly significant group.
UK Business Breach Statistics 2025/26: How Many Firms Were Attacked
UK BUSINESSES REPORTING A BREACH OR ATTACK, BY SIZE (%) (1 block = 2 percentage points)
Micro (1 to 9) | █████████████████████ 42%
Small (10 to 49) | ███████████████████████ 46%
Medium (50 to 249) | ████████████████████████████████▌ 65%
Large (250+) | ██████████████████████████████████▌ 69%
| Measure (businesses) | 2025/26 | Earlier Comparison |
|---|---|---|
| Any breach or attack | 43% (about 612,000) | 50% in 2023/24 |
| Phishing | 38% | 42% in 2023/24 |
| Impersonation | 12% | 17% in 2023 |
| Ransomware | 1% | 3% in 2024/25 and 2023/24 |
| Victim of any cyber crime | 19% (about 267,000) | 22% in 2023/24 |
| Cyber crimes against businesses (estimate) | About 5.19 million | Not applicable |
| Non-phishing cyber crimes (estimate) | About 70,000 | Not applicable |
| Cyber-facilitated fraud victims | 3% (about 43,000) | Not applicable |
| Charities reporting a breach or attack | 28% (about 57,000) | Not applicable |
Source: DSIT and Home Office, Cyber Security Breaches Survey 2025/2026, published April 30, 2026. Survey of 2,112 businesses and 1,085 charities, fieldwork August to December 2025.
The official breach rate has stopped falling. Prevalence held at 43% after dropping from 50% in 2023/24, and medium and large firms are far likelier to be hit than micro firms. Phishing remains the dominant attack at 38%, and among firms that were attacked, 51% experienced phishing only, up from 45%. Among businesses that were victims of cyber crime, 93% had experienced phishing.
Most attacks are low cost for the median firm, but a minority are expensive. The median reported cost of the most disruptive breach was £0, while the top 5% of cases cost £4,000 or more, rising to £10,000 for medium and large firms. The survey only counts attacks that firms identify and are willing to report, so it likely understates the true total.
Ransomware in the UK 2026: Costs of the Biggest Attacks
ESTIMATED COST OF MAJOR UK CYBER ATTACKS, 2025 (£ MILLIONS) (1 block = £50 million)
Jaguar Land Rover (CMC, UK economy) | ██████████████████████████████████████ ~£1,900M
Marks & Spencer (reported) | ██████ ~£300M
JLR direct cost (company) | ███▌ £196M
| Incident | Date | Reported Impact |
|---|---|---|
| Marks & Spencer | April 2025 | Cost reported at about £300 million, with about £100 million insured |
| Co-op Group | April 2025 | Data of 6.5 million members stolen |
| M&S and Co-op combined (CMC) | 2025 | £270 million to £440 million, Category 2 |
| Jaguar Land Rover | Aug 31, 2025 | About £1.9 billion to the UK economy (range £1.6 to £2.1 billion); production halted for about 5 weeks |
| JLR supply chain | 2025 | About 5,000 businesses affected |
| JLR direct cost to company | 2025 | £196 million |
| Ransomware prevalence among UK businesses, 2025/26 | Survey | 1%, down from 3% |
Source: Cyber Monitoring Centre; LBC and Slaughter and May summaries; Farrer & Co; company disclosures as reported.
JLR is the most expensive cyber event on record in the UK, according to the Cyber Monitoring Centre, an independent non-profit that classifies major incidents. Its £1.9 billion estimate includes the manufacturer’s own losses and losses in its supply chain. The group that claimed the attack also claimed the 2025 retail attacks on M&S, Co-op and Harrods, according to reporting.
The government plans to ban ransomware payments by public sector bodies and CNI operators and to require other organisations to notify it before paying. The survey’s 1% ransomware rate looks low, but it measures the share of firms hit, not severity, and a few large attacks drove most of the harm. The sources reviewed do not describe these 2025 attacks as AI-malware attacks.
AI-Enabled Malware: What Google Has Documented
FRONTIER AI CYBER CAPABILITY DOUBLING TIME (MONTHS) (1 block = 1 month)
Previous estimate (AISI) | ████████ 8 months
Current estimate (AISI) | ████ 4 months
| Malware | Type | Status |
|---|---|---|
| PROMPTFLUX | VBScript dropper that asks an LLM to rewrite its own code | Experimental; found June 2025 |
| PROMPTLOCK | Cross-platform ransomware that generates scripts at runtime | Proof of concept |
| PROMPTSTEAL (LAMEHUG) | Data-theft malware that queries an LLM for commands | Used by APT28 against Ukraine; first LLM-querying malware seen in live operations |
| FRUITSHELL | PowerShell reverse shell with prompts meant to evade AI-based defenses | Observed in the wild |
| QUIETVAULT | Credential stealer for GitHub and NPM tokens that uses AI tools | Observed |
| PROMPTSPY | Android backdoor with an AI automation module | Identified by ESET; detailed by Google, May 2026 |
Source: Google Threat Intelligence Group, November 2025 and September 8, 2026 reporting; BleepingComputer; Security Affairs; Fox News (CyberGuy), September 2026. Global findings, not UK-specific.
Google found five malware families with novel AI capabilities in 2025, which it described as the first malware to call an AI model during execution. PROMPTFLUX was experimental and caused no real damage before Google disabled its access, while PROMPTSTEAL was used by a Russian state-linked group in live operations. These are global findings, and the sources reviewed do not give a UK victim count for any of them.
The UK government expects capability to keep rising. Its AI Security Institute assesses that frontier AI cyber capability is now doubling every 4 months, down from 8 months, based partly on its testing of Anthropic’s Mythos model, which it found substantially more capable at cyber offence than earlier models it assessed. In September 2026 Google reported attackers moving from basic prompting toward agentic AI workflows and AI-enabled automation.
UK Government Assessments of AI Cyber Threat: NCSC and the April 2026 Open Letter
| Assessment or measure | Finding |
|---|---|
| NCSC, impact of AI on cyber threat to 2027 | AI will “almost certainly” increase the frequency and intensity of cyber intrusions |
| NCSC, same assessment | Growing AI use in CNI almost certainly widens the attack surface |
| NCSC, same assessment | A “digital divide” will grow between organisations that keep pace with AI-enabled threats and those that do not |
| NCSC, same assessment | AI will highly likely help skilled actors who can fine-tune models to find and exploit zero-day vulnerabilities by 2027 |
| NCSC, CEO speech, June 2026 | By 2028 attackers will probably use AI to exploit known legacy vulnerabilities at scale across CNI |
| Open letter, April 15, 2026 | Ministers urge boards to adopt Cyber Essentials, the Cyber Governance Code and NCSC Early Warning |
| Breaches Survey, 2025/26 | 31% of businesses use, adopt or consider AI; 24% of them have AI security practices |
Source: NCSC, Impact of AI on cyber threat from now to 2027; NCSC CEO RUSI speech, June 17, 2026; UK Government open letter, April 15, 2026; DSIT Cyber Security Breaches Survey 2025/2026.
The NCSC’s central judgement is that AI will help attackers more than defenders in the near term unless organisations keep pace. It expects human-and-machine teaming to speed up vulnerability discovery, exploitation and malware changes meant to evade detection. Its language is probabilistic, such as “almost certainly” and “highly likely,” so these are assessments and not measured counts.
Readiness is limited. Only about a quarter of UK organisations using or considering AI report security practices to manage its risks. The government’s advice in the open letter is the same basic hygiene it gives for older threats: board oversight, Cyber Essentials certification, and the NCSC’s free Early Warning service.
Critical Infrastructure Cyber Attacks in the UK 2026: State-Linked Threats
SHARE OF UK CNI ATTACKS LINKED TO HOSTILE STATES, YEAR TO MAY 2026 (%) (1 block = 3 percentage points)
Linked to hostile states | █████████████████████████ ~75%
Other actors | ████████ ~25%
| Measure | Figure |
|---|---|
| CNI incidents handled by NCSC, year to May 2026 | More than 200 (Computer Weekly reported 200 from January to May 2026) |
| Share linked to hostile state actors | About 75% |
| NCSC Category 2 (highly significant) incidents, Sept 2024 to Aug 2025 | 18 |
| NCSC Category 1 (national cyber emergency) incidents | 0 |
| NCSC assessment of AI exploitation of legacy CNI technology | At scale by 2028 (probable) |
| Cyber Security and Resilience Bill | Before Parliament in April 2026; would strengthen protections for services such as the NHS and energy |
| Ransomware payment ban (proposed) | Public sector and CNI operators |
Source: NCSC CEO Richard Horne, RUSI Annual Security Lecture, June 17, 2026, as reported by UKAuthority, Dig.Watch and Computer Weekly; NCSC Annual Review 2025; UK Government open letter, April 15, 2026.
NCSC chief executive Richard Horne said in June 2026 that about three-quarters of attacks on UK critical national infrastructure were linked to hostile state actors. Reporting differs on the period: UKAuthority and Dig.Watch say more than 200 incidents in the year to May 2026, while Computer Weekly reports 200 between January and May. The NCSC has not yet published the underlying breakdown.
The most severe category has not been reached. The UK has recorded no national cyber emergency, and the NCSC’s 18 highly significant incidents in 2024-25 involved potential, not confirmed, serious disruption to essential services. The government’s Cyber Security and Resilience Bill and the planned National Cyber Action Plan are the main policy responses.
UK Cyber Hygiene Gaps: Controls Organisations Still Lack
UK BUSINESSES WITH EACH CONTROL IN PLACE, 2025/26 (%) (1 block = 2 percentage points)
Up-to-date malware protection | ████████████████████████████████████████▌ 81%
Secure cloud backups | █████████████████████████████████████ 74%
Two-factor authentication | ███████████████████████▌ 47%
Software updates within 14 days | █████████████████ 34%
Formal incident response plan | ████████████▌ 25%
Review immediate suppliers' risk | ███████▌ 15%
Hold Cyber Essentials | ██▌ 5%
| Control (businesses) | 2025/26 | Earlier |
|---|---|---|
| Up-to-date malware protection | 81% | 77% in 2024/25 |
| Backing up data securely via cloud | 74% | 71% |
| Two-factor authentication | 47% | 40% |
| VPN for remote staff | 36% | 31% |
| Policy to apply security updates within 14 days | 34% | Not shown |
| Formal incident response plan | 25% | Broadly stable |
| Cyber insurance of any kind | 47% | 45% |
| Review risks from immediate suppliers | 15% | Not shown |
| Review risks across wider supply chain | 6% | Not shown |
| Hold Cyber Essentials | 5% | 3% |
| Staff training in last 12 months | 19% | 19% |
Source: DSIT and Home Office, Cyber Security Breaches Survey 2025/2026.
Most UK businesses have basic protection but not the controls that stop modern intrusions. Malware protection reached 81% of businesses, yet only 47% use two-factor authentication and 34% have a policy to patch within 14 days. With AI expected to speed up exploitation of known vulnerabilities, slow patching is the gap the NCSC warns about most.
Supply chains and response planning are weak. Only 15% of businesses review risks from immediate suppliers and 6% the wider chain, even though the JLR attack disrupted about 5,000 businesses in its supply chain. Just 25% have a formal incident response plan, and 5% hold Cyber Essentials, up from 3%.
Frequently Asked Questions About AI Malware in UK 2026
How many AI malware attacks happen in the UK?
No official count exists. The NCSC and government surveys do not break out attacks that used AI malware, so figures for all cyber incidents are the best available measure.
How many serious cyber attacks did the UK face?
The NCSC handled 204 nationally significant incidents between September 2024 and August 2025, up from 89, including 18 highly significant incidents.
What percentage of UK businesses were attacked?
43% of UK businesses, about 612,000, reported a breach or attack in the 12 months covered by the 2025/26 survey, and 69% of large businesses.
How common is ransomware in UK businesses?
1% of businesses reported ransomware in 2025/26, down from 3%. A few large attacks, such as JLR and M&S, account for most of the harm.
How much did the Jaguar Land Rover cyber attack cost?
The Cyber Monitoring Centre estimates about £1.9 billion to the UK economy, with a range of £1.6 billion to £2.1 billion. Production stopped for about five weeks.
What is AI-enabled malware?
Malware that uses an AI model while running, for example to rewrite its code or generate commands. Google documented five such families in 2025, including PROMPTFLUX, PROMPTLOCK and PROMPTSTEAL.
Has AI malware been used in live attacks?
Yes. Google reported that Russian group APT28 used PROMPTSTEAL, which queries an AI model, against targets in Ukraine. It was the first such use Google had seen in live operations.
How fast is AI cyber capability improving?
The UK’s AI Security Institute assesses frontier AI cyber capability is doubling every 4 months, down from 8 months previously.
Who is attacking UK critical infrastructure?
The NCSC chief said about 75% of attacks on critical national infrastructure in the past year were linked to hostile state actors, out of more than 200 incidents.
What is the UK ransomware payment ban?
A government proposal to ban public sector bodies and CNI operators from paying ransoms, with a notification scheme for other organisations.
Do UK businesses have protection against AI threats?
Only 24% of UK businesses using or adopting AI report security practices for its risks, and the government advises basic hygiene such as Cyber Essentials.
Is the UK expected to see more AI-driven attacks?
The NCSC says AI will almost certainly increase the frequency and intensity of cyber intrusions through 2027. This is an assessment, not a measured count.
Disclaimer: The data research report we present here is based on information found from various sources. We are not liable for any financial loss, errors, or damages of any kind that may result from the use of the information herein. We acknowledge that though we try to report accurately, we cannot verify the absolute facts of everything that has been represented.
