Nova South building in London, home of the UK National Cyber Security Centre, which issued an AI cybersecurity warning — Nova South in London, where the NCSC launched its Annual Review 2026. Image: NCSC.

AI cybersecurity stopped being a future problem this week. On November 3, the UK's National Cyber Security Centre said AI-powered ransomware is highly likely to hit UK critical infrastructure by 2028.

The warning came in the NCSC Annual Review 2026, which it calls the State of the Cyber Nation report. The NCSC is part of GCHQ, the UK's signals intelligence agency.

I'm Nitin from the Debugged Pro team. The report is about the UK. But the weak spots it lists are the same ones we see in client systems in India and the USA. Here is what was said, and what agencies should do next.

What the NCSC Said About AI Cybersecurity

The NCSC's official announcement is short and direct. These are the main points:

AI ransomware by 2028: It is "highly likely" that AI-powered malware will be used in a ransomware attack on UK critical national infrastructure by 2028.
Faster, bigger attacks: Within two years, hostile use of AI will likely boost the scale and efficiency of ransomware, social engineering and vulnerability exploitation.
A near-term spike: NCSC chief executive Dr Richard Horne warned of "a real prospect of an unprecedented rise in the number of cyber incidents" in the coming months.
China's position: The NCSC says China is highly likely best positioned to use AI against critical infrastructure and other strategic targets.

The review covers the NCSC's tenth year, from September 1, 2025 to August 31, 2026.

"Tech on the Cheap" Is the Real Problem

The most useful part of the report is not about AI at all. It is about old habits.

In his launch speech, Horne said the coming rise in incidents is "not because AI has invented new ways to attack us." It is because long-tolerated weaknesses can now be found and exploited at huge scale.

"For decades we have collectively got away with tech on the cheap," he said. He listed what that looks like in practice:

End-of-life systems kept running "for just one more year".
Upgrades pushed back to save money.
Access granted where it was convenient, not where it was needed.
Services bought only on price, without checking security.
No monitoring to spot when things go wrong.

If you have worked in an agency for a few years, that list probably feels familiar. It certainly does to us.

The Numbers Behind the Warning

The NCSC also shared figures from its incident management team for the year. They help explain why the tone is so urgent.

Bar chart of NCSC incident figures: 8,500 organisations warned, 1,831 reports, 337 incidents supported, 80 nationally significant — NCSC incident figures for September 2025 to August 2026. Source: NCSC Annual Review 2026. Chart: Debugged Pro.

1,831 reports of cyber attacks were received.
337 incidents needed direct NCSC support.
80 of those were classed as nationally significant.
8,500 organisations were warned of a suspected compromise through the Early Warning service.

One detail stands out. The team has started to see attackers using AI tools. But none of the observed uses enabled a new class of attack.

Most successful attacks still used familiar methods. Known flaws in internet-facing systems, stolen passwords and social engineering. Many state-linked hacks used flaws that already had patches which simply were not installed in time.

How AI Changes the Attack, in Plain Words

The NCSC's assessment of AI and cyber threats makes a few clear judgements. Here they are in everyday language.

AI Is a Speed Multiplier

The NCSC says AI is almost certain to act as a force multiplier. It helps with scouting targets, social engineering, finding flaws and writing exploit code.

Skilled attackers get the biggest boost. Human operators are still needed to direct and check the AI's work.

The Patch Race Gets Shorter

AI models can now spot flaws in large codebases that went unnoticed for years. The NCSC expects defenders to face a race to patch before attackers build working exploits.

The Guardian's report gave a recent example. Last month, AI helped uncover a critical weakness in PHP, a language the paper says is used by almost 70% of websites. Horne said the NCSC sent over 110,000 notifications and early warnings to UK organisations about that PHP flaw.

Shadow AI Is a Data Leak Waiting to Happen

The NCSC also flagged "shadow AI". That is staff using unapproved AI tools, often by moving work data onto personal devices.

When that happens, a company loses sight of where its data goes and who can see it. For agencies holding client code and customer data, this one deserves attention.

A Growing Gap Between Prepared and Unprepared Firms

The report expects a widening divide. Firms with mature security and good AI governance will defend better. Firms with old tech and thin budgets will become easier, more attractive targets.

Real-World Signs It Is Already Happening

The Guardian's Dan Milmo listed some of the targets in scope. The UK's critical infrastructure covers energy, transport, defence, food, health and finance.

Horne also pointed to recent attacks on supermarket refrigeration in Europe. The Guardian reported that hackers pushed fridge temperatures above 50C.

We covered a related case recently. Our breakdown of the ARTEX AI pentesting agent used in real attacks shows how AI agents are already chaining together scanning, exploits and stolen credentials.

What This AI Cybersecurity Warning Means for Agencies in India and the USA

You might think a UK report is not your concern. I'd push back on that.

The weak spots the NCSC lists are not British. Old WordPress plugins, unpatched servers, shared admin passwords and forgotten staging sites exist everywhere. AI just makes them faster to find.

Agencies sit in an awkward spot. We often hold keys to many client systems at once. That makes us part of their supply chain, and the NCSC calls out supply chain risk directly.

For Agencies Serving Indian Clients

Many Indian SMEs run on tight IT budgets. That is exactly the "tech on the cheap" pattern the NCSC describes.

Keep in mind that CERT-In rules require many cyber incidents to be reported within six hours. A clear incident plan is no longer optional.

For Agencies Serving US Clients

US clients often ask about compliance first. A good starting point is CISA's Known Exploited Vulnerabilities catalog, a free list of flaws attackers are using right now.

If a client system has anything on that list, patch it before you build anything new.

A Practical AI Cybersecurity Checklist for Your Agency

None of this needs a huge budget. It needs discipline. Here is the plan we use with our own project teams.

Diagram pairing NCSC AI cybersecurity judgements with practical actions such as faster patching, passkeys and backup testing — NCSC judgements on AI threats, paired with simple actions. Source: NCSC Annual Review 2026. Diagram: Debugged Pro.

Build an inventory. List every server, domain, plugin, API key and AI tool across your client projects. You can't protect what you don't know exists.
Retire end-of-life tech. Flag anything out of vendor support. Give clients a written upgrade plan with dates and costs.
Shorten patch times. Set a target, such as critical patches within days, not weeks. Track it every month.
Take things off the internet. Admin panels, databases and old staging sites rarely need to be public. Put them behind a VPN or remove them.
Switch to passkeys. The NCSC wants passkeys to be the norm. Start with admin, hosting and email accounts.
Set AI tool rules. Publish a short list of approved AI tools. Say clearly which client data must never go into them.
Turn on monitoring. Basic logging and alerts on login failures and unusual traffic catch many attacks early.
Test your backups. A backup you have never restored is a guess. Run a restore drill every quarter.

Use AI on Defence Too

The NCSC is clear that AI also helps defenders. Horne noted the Linux Kernel Project patched more than a thousand vulnerabilities last month.

The UK is backing this with money too. That includes £60 million for the first phase of a national Cyber Shield. Another £70 million goes to a new National Exposure and Real-Time Vulnerability Centre.

For agencies, a simple start is AI-assisted code review and dependency scanning on every pull request. Just keep a human in charge of what gets merged.

Where to Get Help

If your team needs support with AI governance, risk reviews or secure AI rollouts, our AI security, risk and compliance services cover that work. For legacy systems that need a planned upgrade, our software development consulting team can map a realistic path.

What We Don't Know Yet

The NCSC has not named a specific AI ransomware strain targeting infrastructure today.
Its 2028 scenarios are assessments, not firm predictions.
Details of the UK's upcoming National Cyber Action Plan have not been published.

Key Takeaways

The NCSC expects AI-powered ransomware to hit UK critical infrastructure by 2028.
AI is not inventing new attacks. It is exposing old, unpatched weaknesses at scale.
Most breaches still use familiar methods, like known flaws, stolen passwords and social engineering.
Shadow AI and slow patching are two of the biggest risks for agencies.
AI cybersecurity starts with basics: inventory, patching, passkeys, monitoring and tested backups.

Frequently Asked Questions

What did the NCSC Annual Review 2026 say about AI?

It said AI-powered malware is highly likely to be used in a ransomware attack on UK critical infrastructure by 2028.

Is AI creating new kinds of cyber attacks?

Not yet, says the NCSC. AI is mainly making existing attacks faster and bigger.

Why should agencies in India and the USA care?

The same weak spots exist in client systems everywhere. AI-powered attackers do not stop at borders.

What is shadow AI?

It is staff using unapproved AI tools, which can leave company data exposed.

What is the first AI cybersecurity step for a small agency?

Make an inventory of systems and AI tools, then patch or retire anything out of support.

About the Author

Nitin Kumar is the Lead Project Coordinator at Debugged Pro. He works with AI and software project teams serving clients in India and the USA. He writes about practical ways agencies can use new technology safely and without cutting corners.