VULNpocalypse
InformationalVulnerabilityMonitoring

NCSC warns AI-powered ransomware is highly likely to hit UK critical infrastructure by 2028

The UK NCSC says AI-powered malware is highly likely to feature in ransomware attacks on critical infrastructure by 2028, so patch and fix the basics now.

Published
Exploitation
No confirmed AI-powered ransomware attack on UK critical infrastructure yet, the NCSC statement is a forecast, and responders have seen attackers use AI tools but none enabled a new class of attack

What happened

On 3 November 2026 the head of the UK National Cyber Security Centre (NCSC), Richard Horne, launched the agency’s annual report and warned of “a real prospect of an unprecedented rise in the number of cyber incidents” in the coming months (Infosecurity Magazine, The Record).

Key points from the coverage:

  • The NCSC says AI-powered malware is “highly likely” to feature in a ransomware attack on UK critical national infrastructure by 2028. It also expects China, Russia and other states to routinely use AI in their cyber operations by then (The Record, Cyber Security News).
  • Infosecurity Magazine reports the report names financially motivated attackers and targets such as power stations, airports and hospitals (Infosecurity Magazine).
  • This is a forecast, not confirmation that such an incident has happened (Cyber Security News).
  • Horne said the risk is “not because AI has invented new ways to attack us” but because long-tolerated vulnerabilities can now be found and exploited at scale (The Record).
  • The report says AI is more likely to enhance and scale existing tactics than to find brand new vulnerabilities (SecurityWeek).
  • NCSC responders have begun seeing attackers use AI tools, but none of the observed uses enabled a new class of attack (Cyber Security News).
  • Officials say experts have not yet observed a corresponding rise in exploitation. They point to vendors fixing hundreds of vulnerabilities in a single update, and to the Linux kernel project patching more than a thousand vulnerabilities in the previous month (The Record).
  • Even state-linked intrusions often went through internet-facing applications, exploiting vulnerabilities for which patches existed but had not been deployed in time (The Record).

Figures from the annual review, covering 1 September 2025 to 31 August 2026 (Cyber Security News):

  • 1,831 attack reports.
  • 337 incidents needing NCSC support, of which 80 were nationally significant.
  • 8,500 organizations notified through the Early Warning service.
  • 88% of incidents across large organizations involved four initial access routes: public-facing applications, phishing, valid accounts and external remote services.

The NCSC cautions against reading incident totals as a simple measure of rising or falling threat.

Other details from the coverage:

  • Horne cited the PHP2Shell vulnerabilities as the best recent example. The NCSC issued 110,000 proactive notifications and Early Warning alerts to UK organizations over them (Infosecurity Magazine). Technical details of those flaws are not confirmed in these sources.
  • SecurityWeek separately reports that last week two critical flaws found with AI assistance in a programming language used by 70% of websites led to urgent security updates. It does not name the language or give identifiers, so any link to PHP2Shell is not confirmed (SecurityWeek).
  • The Register reports that the NCSC alluded to open source AI security tools being repurposed for abuse. It says researchers see widespread abuse of ARTEX, an open source, Chinese-made AI red-teaming platform, against targets including UK registrar 123 Reg and Microsoft 365 since July (The Register).
  • Government funding announced includes £60 million for the first phase of Cyber Shield, £70 million for the NERV Centre, £40 million for a National Agentic Security Unit and £24 million for public-sector defenses (Infosecurity Magazine, The Register). Project DROMOS, which started as a four-week pilot, found hundreds of potential security issues including critical vulnerabilities (The Record).
  • The government also announced a 75% discount on Cyber Essentials certification for micro, small and medium-sized businesses (Infosecurity Magazine).

Am I affected?

This is a threat assessment, not a single product flaw. No CVE, affected version or fixed version is given in the sources, so those details are not confirmed.

You are most exposed if any of these apply:

  • You operate UK critical national infrastructure or supply it. The NCSC names power, airports and hospitals, and Horne cited electricity switching, water sluice gates, dams, manufacturing lines and refrigeration controls (Infosecurity Magazine, SecurityWeek).
  • You run public-facing applications, external remote services or accounts with weak protection, which are the access routes in 88% of the large-organization incidents the NCSC handled (Cyber Security News).
  • You keep end-of-life systems running or delay upgrades for cost reasons, which Horne called “tech on the cheap” (The Record).
  • Your staff use unapproved AI services, which can expose sensitive company information (Cyber Security News).
  • You run PHP-based web applications, given the PHP2Shell notifications. Exact exposure criteria are not confirmed.

What to do now

  1. Build an accurate inventory of systems and identify everything internet-facing, including remote access and PHP-based sites (Cyber Security News).
  2. Patch exposed services first. The NCSC notes many state-linked intrusions used available patches that were not deployed in time (The Record).
  3. Prepare for a “patch wave” by improving your ability to deploy fixes quickly, frequently and across suppliers. Not every AI-discovered flaw will be exploitable, but delays give attackers openings (Cyber Security News).
  4. Replace or retire end-of-life systems and stop deferring upgrades (Infosecurity Magazine).
  5. Limit administrator access, control account access and consider wider passkey adoption (SecurityWeek, Cyber Security News).
  6. Back up important files and data, protect those backups, and rehearse incident response (SecurityWeek, Cyber Security News).
  7. Map supplier dependencies and treat Cyber Essentials as the starting point, scaling up resilience with your size, complexity and risk (Cyber Security News, The Register).
  8. Set a policy for approved AI services so sensitive data is not sent to tools you cannot track. If you adopt AI-driven defensive automation, give it clear limits and human oversight (Cyber Security News).
  9. Watch for abuse of dual-use AI red-teaming tools such as ARTEX against your Microsoft 365 and other exposed services. Specific indicators are not confirmed in these sources (The Register).

Sources