UK cyber chief says old weaknesses, not AI, are the real risk
The UK's NCSC warns AI is speeding up attacks on long-ignored security gaps and supply chains. Here's what IT teams should take from it.

The UK's National Cyber Security Centre (NCSC) isn't saying AI has invented a new kind of attack. Its message is harder to hear: the attacks AI is making cheaper and faster are exploiting weaknesses that organisations have tolerated for years. Announcing its latest annual report on the state of UK cybersecurity, NCSC CEO Richard Horne compared the country's security posture to the Leaning Tower of Pisa, a structure that has stood for centuries on flawed foundations.

If you run infrastructure, that framing should land. The problem isn't only the new tooling on the attacker side. It's the patch backlog, the unmanaged legacy systems and the suppliers nobody has mapped properly.
What the NCSC report found
The agency received 1,831 reports of cyberattacks this year. Of those, 80 were classified as nationally significant. It also notified 8,500 organisations of suspected attacks. Those numbers describe a reporting and notification workload, not a full count of incidents across the economy, so treat them as a snapshot of what reached the NCSC.

The report expects AI to be built further into threat actors' operations, raising the speed, scale and efficiency of their campaigns. It also names China's adversarial cyber activity as the biggest "cyber challenge" facing the UK. Horne said the period of post-Cold War stability has ended, and pointed to sustained pressure on military networks, Russian campaigns that stay below the threshold of armed conflict, and continuing Iranian cyber operations despite the wider Middle East conflict.
Horne's warning was direct: "In the coming months, we face a real prospect of an unprecedented rise in the number of cyber incidents." That's a forecast from the agency responsible for national cyber defence, and it should shape how you plan your next few quarters.

Why old problems are the real exposure
Here's the thing: AI lowers the cost of finding and exploiting known weaknesses. It doesn't need to discover anything new to be dangerous. Horne put it plainly: vulnerabilities that have been tolerated for far too long can now be found and exploited at a scale and pace that was once unimaginable.
He also pointed at supply chains. Many organisations, he said, rest on weak technology and operational foundations, and those foundations extend beyond their own walls. A well-hardened internal estate still inherits risk from a vendor, a managed service provider or a shared platform that hasn't kept pace.

That's the uncomfortable part for MSPs and service desks. A client's security posture is only as strong as the weakest link they connect to, and attackers have every reason to go looking for that link instead of the front door.
Recent incidents that fit the pattern
The NCSC's warning arrives against a busy backdrop. A second wave of attacks hit retail units, and grocery stores across Ireland, the UK, Belgium and Sweden were targeted earlier in the week, causing food waste and losses reported in the hundreds of thousands. A fresh ransomware attack on a school estate in the UK was reported on 3 November. Last week, suspected North Korean hackers reportedly stole more than $6bn in cryptocurrency assets from Binance, which would make it the biggest bank heist in history.

Those incidents cover very different sectors, but the operational lesson is similar: distributed estates, shared systems and dependencies that are hard to see from the middle of the network are where the damage tends to start and spread.
A recent Eurobarometer survey found that three in four EU-based employees faced potential phishing or hacking attempts in the past six months. That's a reminder that social engineering still reaches the people your technical controls depend on, and that AI-generated lures are only going to make that harder to spot.

Industry response and a worrying gap
AI developers are part of the response. The NCSC's context notes that Anthropic and OpenAI, creators of many of the leading AI models, have teamed up with organisations to help secure their systems in what has become a high-stakes race. Across the board, though, businesses and government departments still appear to have a long way to go.
The gap is most visible in spending. Several Irish businesses have reportedly cut their cybersecurity budgets despite rising risks, a move one commentator described as a "high-risk strategy". When threat activity is climbing and the cost of an incident is rising with it, trimming security spend is a bet that nothing will test the cuts. Teams in that position often find the budget question returns after a breach, at a much higher price.

Image: © norbel/Stock.adobe.com
What IT teams should be checking now
The NCSC report doesn't include a technical checklist, so treat what follows as general framing rather than official guidance. Still, the report's logic points to a few places to look.
Start with your supply chain map. If you can't list the third-party platforms, remote management tools and managed services that touch your environment, you can't judge which of them would hurt you most if compromised. Next, look at your patch and vulnerability backlog. Horne's point about tolerated vulnerabilities suggests the old items on that list deserve as much attention as the new ones. Review recovery readiness too: a recent report covered in the same coverage stream found that recovery readiness is a missing link in cyber resilience, and the ransomware incident on a school estate shows why restoring data and services matters as much as stopping the initial intrusion.
Finally, check whether your security budget reflects the threat trend the NCSC describes. If the forecast of an unprecedented rise in incidents holds, the cost of thin defences will show up sooner than most planning cycles assume.
My read is that the next year will reward teams that boring-but-essential work is done well: asset visibility, timely patching, tested backups and clear accountability for suppliers. AI will keep making attackers faster. The organisations that cope best will be the ones that stopped treating their known weaknesses as an acceptable risk.