The UK government has issued a timetable in what has been something of a speculative and unsure topic so far: it is warning the nation’s critical infrastructure organizations that they can expect AI-powered malware attempts by 2028, with an “unprecedented rise” in the number of cyber incidents as AI tools become more broadly available to threat actors and accelerate the capabilities of the most skilled and well-funded.
As highly capable frontier AI models have become available throughout 2026, there has already been some observed adoption of them by threat actors. Harnessing these advanced capabilities has mostly been limited to more advanced groups, however, and serious incidents remain few and far between at present; the UK government’s central warning is that the barrier of entry to use of these tools will be sharply reduced throughout 2027 into 2028, and organizations can expect regular assaults on any vulnerabilities that have not yet been discovered and patched.
UK government anticipates near-term AI-powered malware spike
The warning comes as part of the National Cyber Security Centre (NCSC)’s “Annual Review” for 2026, also referred to as the “State of the Cyber Nation” report. The introduction opens with a warning that the cyber threat landscape is changing fast, though the change is not so much brand new developments in threat vectors but the fact that malicious AI tooling is likely to become more commonly available (and used skillfully by advanced threat actors) in a short period of time.
As the report notes, the central threat is not that AI is creating new vulnerabilities but that it is very rapidly revealing existing ones that are owed to years of technical debt. This has created a massive workload to catch up with, but one that responsible use of defensive AI can also do a great deal to reduce. The report also notes that combating AI does not require much new beyond that; it is down to the fundamentals that already exist like timely patching, cyber literacy and teaching security hygiene.
Still, there will remain holes and gaps to exploit in the era of persistent AI discovery. The big problem is that threat actors will be both more numerous and more efficient and effective, thanks to availability of AI tools. The report notes that Chinese state-backed hackers and other advanced threats in that uppermost tier will use AI as a force multiplier and present the greatest risk, with AI providing more benefit the more skilled the human operator is. But it will also considerably reduce the barrier of entry to newer players, and automated AI exploitation of vulnerabilities and development of malware will help the smaller players become bigger and more persistent threats.
Darin Fredde, Sr. Director of Technical Marketing Engineering at Ridge Security, notes that this is a separate threat from fully automated AI as it will be human-driven: “We should be careful not to confuse AI-assisted attacks with fully autonomous threats. We should also recognize that AI offers defenders opportunities to discover vulnerabilities, correlate evidence, and prioritize remediation. But capability is different from verification. An AI system reporting that a vulnerability has been fixed is not proof that the attack path has been eliminated.”
“The challenge is to use automation to continuously test our assumptions, independently validate outcomes, and strengthen our ability to contain and recover from disruption,” Fredde adds. “The question isn’t whether attackers will get faster. It’s whether we can make verified resilience move faster than the threats we’re preparing to face.”
The report specifically warns that UK critical infrastructure will be “highly likely” to be hit by an AI malware attack that involves ransomware by 2028. It notes that cyber criminals are the particular risk in this area; save for odd exceptions like North Korea, state-backed actors generally prefer low-key espionage approaches and do not engage in brazen ransomware attacks. This will likely be a result of the cost of operations dropping for these profit-seeking actors, combined with greater ease in locating unaddressed vulnerabilities, all of this made possible by AI assistance.
Cyber Shield Program requires new AI defense integration and automated testing for critical infrastructure in 2027
The report notes that jailbroken AI models are already available to “street level” criminal threat actors. One specific source that has become something of a central clearinghouse is the “Kriminal.ai” service, a clear web service that sells jailbroken access to leading models for a flat subscription fee of $13 per month. Independent criminal groups have also already pulled off several notable attacks on critical infrastructure and government agencies using AI tools, most notably the broad breach of the Mexican government in late 2025 that emerged even before the more dangerous frontier models introduced in 2026 were available.
Joshua Marpet, Senior Product Security Consultant at Finite State, provides some perspective on the expected scope and cost for smaller players to increase their capability: “As AI tokens are still fairly inexpensive, and even a small inference infrastructure can create phishing emails that pluck heartstrings or push a false sense of urgency, it’s relatively trivial to build that route into a company or agency. Once you build the way in, getting someone to click a malicious link is pretty easy.”
“This is not “huge ai writes incredibly complex formerly unseen zero-day vulnerability!!!,” Marpet notes. “This is “tiny AI writes an email to elicit an emotional reaction of sorrow, pity, or urgency”. So rather than spending $10k to build a new zero day, I can spend 20 minutes to write an email and some shell code. For free. From hugging face or evilgpt. But if you feel the need to build some revolutionary malware, you can always spend about 120k and build your own high end ai infrastructure, with a free Chinese model such as Kimi, Qwen, or deepseek. 120k?!?? One or two good hits on the ransomware, and you’ll make your money back.”
The UK government thinks such an attack on its critical infrastructure is inevitable by 2028 due to the current pace of growth of cloud controlled CNI, greater integration of cyber into physical labs and proliferation of AI-controlled and internet-connected robots and devices. However, in a speech that accompanied the release of the report, NCSC chief executive Richard Horne warned the general public that an unprecedented wave of cyber attacks on critical infrastructure is “imminent.” The UK, like numerous other national governments as of late, has been urging citizens to stock up on days worth of bottled water and non-perishable foods in the event cyber attacks caused extended utility disruption.
The Cyber Security and Resilience Bill does set some new requirements for critical infrastructure companies in 2027 as well, aimed at more forcibly improving resilience against AI malware and other possibilities of this sort. Existing regulations expand to cover some previously ineligible entities, such as electricity load controllers (such as charging stations) and managed service providers, and new supply chain security monitoring requirements kick in. Enforcement penalties are also set to increase substantially.
