On this page (7 sections)
The UK National Cyber Security Centre (NCSC) has released a report called Artificial intelligence and cyber intrusion, listed with a date of 2026-11-03. It looks at how AI is shaping cyber threats today and sketches three possible threat landscapes for 2028.
This is a forward-looking assessment, not a vulnerability disclosure or an incident report. Site owners, agencies and hosting administrators should care because of one recurring theme: if weaknesses are found and used faster, slow patching and loose configuration cost more.
What the NCSC published
The NCSC presents the document as an examination of present AI-enabled threat trends combined with projections to 2028. It says the scenarios are plausible medium-term possibilities and that they are not predictions. It also warns that AI capability and its use in cyber operations are moving quickly, so the chance of surprise is high.
The three scenarios it works through are:
- Baseline: the global threat from AI-cyber capability grows.
- Pessimistic: AI-cyber capability spreads much faster than the UK can defend against it.
- Optimistic: AI-cyber development slows, leaving the UK time to build solid countermeasures.
Every judgement in the report is expressed with the PHIA probability yardstick, where terms such as "likely", "highly likely" and "almost certain" map to defined likelihood ranges. The text we had access to does not list those ranges, so read the wording as calibrated judgement rather than certainty.
Key judgements in plain language
The NCSC's headline judgements look ahead to 2028:
- Highly likely: Russia, China and other nation states will routinely use AI-cyber to pursue geopolitical goals, and the UK will face sophisticated attacks at greater scale and speed.
- Almost certain: many systems will still be open to AI-cyber tools that find and exploit weaknesses. The NCSC says this lets attackers worry less about losing a foothold, since replacing it costs little.
- Likely: attackers will point AI-cyber capability at known vulnerabilities in legacy critical national infrastructure (CNI) technology, and at the new attack surface that comes from deploying AI tools inside CNI.
- Highly likely: AI-powered malware will have been used in a ransomware attack on UK CNI at some point between now and 2028.
The baseline scenario: faster discovery, faster patching
In the baseline picture, the NCSC says widely available AI-powered cyber tools have almost certainly driven a significant rise in zero-day discovery. It also describes defensive AI tools narrowing the gap between a zero-day and an n-day vulnerability, which makes patching quicker.
The caution is the part site owners should note. The NCSC says that unless this defensive technology is configured correctly and widely, many systems stay vulnerable. It adds that an organisation's own patching and testing processes must evolve to match the faster pace of vulnerability discovery.
On state actors, the report says Russian cyber actors have highly likely adapted and retrained open-source AI models instead of building capability from scratch, and that this has highly likely increased pre-positioning, espionage and intellectual property theft aimed at UK interests. Only one independent publisher, the NCSC, covers this story so far, so none of these judgements has been confirmed by a second source in our research.
What this means for website owners
Nothing in the report says a particular plugin, server or CMS is at risk today. The useful lesson is about tempo. If finding and exploiting flaws gets faster, the time between a fix being published and an attacker using the flaw shrinks, and a slow update cycle becomes your biggest exposure.
By that logic, the most exposed sites run outdated software, abandoned plugins or themes, legacy systems nobody owns, or servers with default or half-finished security settings. Agencies and hosts responsible for many sites carry that risk many times over.
What to do now
These steps are general hardening practice that fits the NCSC's focus on patching speed and configuration. The report does not list them as requirements.
- Inventory every site, plugin, theme, library and server you run, including old staging copies and forgotten projects.
- Set a target time for applying security updates, then rehearse your process to confirm it can meet that target.
- Turn on automatic security updates where your stack allows it safely, and keep a tested rollback path.
- Remove or replace software that is no longer maintained, especially legacy components you cannot patch.
- Review the configuration of servers, admin panels and any AI tools you have deployed, since the NCSC ties poor configuration to lingering exposure.
- Use multi-factor authentication on hosting, CMS and registrar accounts, and keep offline or immutable backups in case of ransomware.
- Subscribe to security advisories for the software you depend on so you hear about fixes on release day.
To see which WordPress plugins have updates waiting, run this read-only WP-CLI command from the site's root directory. Any plugin it lists has an update available.
wp plugin list --update=availableOn a Debian or Ubuntu server, this read-only command shows packages with upgrades waiting, which tells you how far behind the operating system is.
apt list --upgradableMonitoring after you patch
Quicker attacks reward quicker detection. Watch for unexpected file changes, new admin users, new scripts on checkout pages, DNS changes and certificate changes. Webforta Monitor covers uptime, certificates, DNS and header changes, and the free security headers checker shows how your response headers are set. Neither replaces patching, and no tool makes a site fully secure.
For related reading, see our earlier coverage: NCSC: AI-Powered Ransomware Warning for UK Infrastructure.
What remains uncertain
- The scenarios are explicitly not predictions, and the NCSC says surprise is highly possible.
- Our research rests on a partial excerpt of one primary source, with no independent confirmation yet.
- The report assesses the UK and CNI; how its judgements apply to individual websites is our inference.
- The excerpt gives no CVE identifiers, severity scores or specific affected products.
Sources
This article is based on the public sources below. Details can change as vendors and researchers publish more; check the original advisories before acting on version numbers.
