The UK’s National Cyber Security Centre, part of GCHQ, published its Annual Review 2026 on 3 November. It says: “It is also highly likely that between now and 2028, AI-powered malware will have been used in a ransomware attack on UK CNI”, meaning critical national infrastructure. The Guardian reported the warning first. On the probability yardstick used across UK intelligence assessment, “Highly Likely” covers a range of about 80% to 90%.
The same review counts 80 nationally significant incidents in the year to August 2026, against 204 a year earlier: a 61% fall, by our arithmetic. They were 24% of the 337 incidents the NCSC handled, down from 48% of 429, also by our arithmetic. The review cites UK government research that malicious cyber activity costs UK organisations an estimated £14.7 billion every year.
Business Pill · Least Privilege
An evergreen Business Engineer explainer on least-privilege access, giving a system only the access it needs. It teaches the general idea only and does not discuss the NCSC review.
The key insight: The NCSC rates an AI-powered ransomware attack on UK critical national infrastructure by 2028 as highly likely, a band of about 80% to 90% on the UK intelligence yardstick, while its own count of nationally significant incidents fell 61% to 80 in the latest year, by our arithmetic.
What the NCSC Assessed
The 2028 judgement sits in a section of the review headed “Technology advances are changing and increasing the threat”. The same passage says “By 2028, it is highly likely China, Russia, and other nation states will routinely use AI cyber to achieve their geopolitical goals with the UK facing an increase in scale and speed of sophisticated cyber attacks”.
On its page on the impact of AI, the NCSC says “It is almost certain AI will act as a force multiplier across the cyber threat landscape”. It adds that “The greatest impact of AI will be the acceleration and scaling of existing tactics, techniques and procedures (TTPs) as opposed to generating new attack vectors”.
On ransomware, the page says criminal use of AI “will highly likely increase the scale and efficiency of ransomware operations, social engineering and vulnerability exploitation”, and that this “is expected to further reduce operational costs for criminals while increasing attack volumes against less resilient targets”.
The NCSC also says “China highly likely has the intent and is best positioned to use AI for operations against critical national infrastructure and other strategic targets”.
Cyber incidents handled by the UK’s National Cyber Security Centre and those it classed as nationally significant, years to August, 2021-22 to 2025-26. Source: NCSC Annual Review 2025 and NCSC Annual Review 2026. Shares and the 61% fall are our arithmetic. The NCSC says reporting does not follow a simple year-on-year trend. Chart: FourWeekMBA.
Animated: incidents the NCSC classed as nationally significant, years to August, 2021-22 to 2025-26: 63, 62, 89, 204 and 80. Source: NCSC Annual Review 2025 and 2026. The 61% fall and the 48% share are our arithmetic. Reports handled by the NCSC, not all UK attacks.
What ‘Highly Likely’ Means
The NCSC says its assessment team uses the probability yardstick of the Professional Head of Intelligence Assessment (PHIA) “every time we make an assessment, judgement, or prediction”. The government’s explainer says it “splits the probability scale into seven ranges”: “≈80% – ≈90%: Highly Likely” and “≈95% – <100%: Almost Certain”, with “≈55% – ≈75%: Likely or Probable” below them.
On that scale, the AI ransomware judgement is one band below the review’s “almost certain” judgement on AI as a force multiplier. The yardstick ranges are not continuous, and the review gives no single percentage of its own.
The Incident Count Went the Other Way
According to the review’s incident page, the NCSC’s incident management team received 1,831 reports of cyber attacks in the year to August 2026. After triage, 337 incidents needed its support, “of which 80 were classed as nationally significant”. Its Early Warning service notified 8,500 organisations of events suggesting a suspected compromise.
Last year’s review put the earlier years at 63, 62, 89 and 204 nationally significant incidents from 2021-22 to 2024-25, out of 355, 371, 430 and 429 handled. By our arithmetic, this year’s 24% share is below 2024-25’s 48% but above the 17% to 21% of the three years before.
The NCSC warns against reading a trend into this: “It is important to note that incident reporting does not follow a simple year-on-year trend”. It says the Early Warning notifications are “likely to have headed off a number of incidents before they developed into events requiring NCSC intervention”. Last year’s review said its data “is not a true reflection of the number of cyber incidents that impact the UK”.
On AI in actual incidents, the team says it “has also begun to see threat actors incorporating AI tools into their attacks”, and that “None of the observed uses of AI enabled a novel class of attack”.
Who Pays for the Defence
The review says the £14.7 billion a year is “equivalent to the size of the UK’s cyber security sector”. Its case study on the 2025 attack on Jaguar Land Rover says the carmaker “reported approximately £196 million in direct cyber-related” costs.
On the public side, the security minister’s foreword says the Government Cyber Action Plan is “backed by more than £210 million of investment”. The NCSC’s release says the Chancellor announced “£60 million into developing the first phase of a national Cyber Shield” and “£70 million for a new National Exposure and Real-Time Vulnerability (NERV) Centre” last week.
The review’s AI page names the split it expects: “It is almost certain a widening digital divide will emerge between organisations able to adopt AI for defence and those unable to adapt”. It says defenders face “a race to shorten patch deployment timelines”.
In a speech launching the report, NCSC chief executive Richard Horne is expected to say, according to the release: “For decades, we have collectively got away with doing tech on the cheap”. The release says he will also say “In the coming months, we face a real prospect of an unprecedented rise in the number of cyber incidents”.
As we read it, the business lever is the cost of standing still: the review’s own judgements put the extra pressure on organisations running legacy, internet-exposed systems, which must patch faster or spend more on defence. That is a reading of the NCSC’s text, not a forecast.
Where Our Coverage Fits
We have tracked AI in attacks this autumn: a China-linked group running AI agents against more than 200 targets, Chainalysis on an agentic attack that exposed customer emails, and Palo Alto’s Arora on the same attacks coming faster. The Guardian reports that Horne will cite the attacks on supermarket fridges in Europe, and that AI helped find a critical weakness in PHP last month; we covered early attacks on PHP and Google’s ban on TV box maker Videostrong over a botnet.
The Structural Read
As we read it, the review separates two things: the incident count it can see, which fell, and a judgement about capability, which hardened. The NCSC itself says the count does not follow a simple year-on-year trend.
As we read it, the review’s cost case rests on old weaknesses rather than new attacks: its incident team says none of the AI uses it observed enabled a novel class of attack.
As we read it, the “digital divide” judgement is a budget question for operators: organisations that can fund AI-assisted defence and faster patching sit on one side of it.
NCSC Annual Review 2026, The impact of AI on cyber threat
“It is almost certain a widening digital divide will emerge between organisations able to adopt AI for defence and those unable to adapt”
Three Implications
FOR INFRASTRUCTURE OPERATORS The NCSC says the extra pressure falls on organisations with legacy technology and limited resources, and that defenders face a race to shorten patch timelines.
FOR THE PUBLIC PURSE The review cites more than £210 million behind the Government Cyber Action Plan, and the NCSC release lists £60 million for a first phase of Cyber Shield and £70 million for the NERV Centre.
FOR READERS TRACKING THE NUMBERS The 80 nationally significant incidents are reports handled by the NCSC, not all UK attacks; the review’s yardstick term covers a probability range, not a point estimate.
The Business Engineer Lens
This story maps onto the Business Engineer framework Inside Anthropic’s Permission Layer.
The analysis describes the governance layer this way: “It is where decisions are made about who gets access to frontier intelligence, under what conditions, and for which purposes.”
As we read it, the NCSC’s divide between organisations that can adopt AI for defence and those that cannot is an access question of that kind; the Guardian quotes Professor Alan Woodward on the need for access to cutting-edge AI. The analysis was published in June 2026 about Anthropic; it does not discuss the NCSC review.
Related coverage: China-linked group ran AI agents against 200+ targets · Freezer cyberattack hits supermarkets in 7 European nations · Chainalysis on an agentic attack
What Is Not Established
The review does not name a sector, a date or a group for the AI ransomware judgement, and gives no percentage of its own beyond the yardstick term. The 2026 pages we read do not give a count of highly significant incidents, which last year’s review put at 18. The incident figures cover reports to the NCSC, not all UK attacks. Horne’s lines are from the NCSC’s release ahead of his speech. We did not contact the NCSC.
This Week In The AI Supercycle
Anthropic’s Reality Check, The Capital Squeeze, & The New World Order
This week’s issue: for two years the AI build-out was priced as a growth story. This week it started being priced as a credit story.
Read the full issue →
Business Engineer Framework
Inside Anthropic’s Permission Layer
A Business Engineer analysis of the governance layer that decides who gets access to frontier AI, under what conditions and for which purposes.
Read the Map of AI →
The Bottom Line
The NCSC rates an AI-powered ransomware attack on UK critical national infrastructure by 2028 as highly likely, the yardstick’s band of about 80% to 90%. Its own count of nationally significant incidents fell 61% in the latest year, to 80, by our arithmetic, and it says that count does not follow a simple trend.
A note on sourcing. The assessments, incident counts and cost figures are from the NCSC Annual Review 2026 (PDF and web pages) and the incident page of the NCSC Annual Review 2025; the probability bands are from the government’s published yardstick; Horne’s lines are from the NCSC’s release ahead of his speech. The Guardian reported the review first. Percentages derived from the counts are our arithmetic. We did not contact the NCSC.
——————————————————–
Click Here For The Original Source.
