NCSC: AI malware likely to hit UK critical infrastructure by 2028

Britain's cyber security agency has warned that malicious software driven by artificial intelligence is highly likely to feature in ransomware strikes on the country's critical services before 2028, according to The Guardian Technology. The National Cyber Security Centre (NCSC) said such an attack on national infrastructure is probable in the period up to that year.

The NCSC, which sits within the GCHQ intelligence organisation, made the assessment in its annual review. It also cautioned that Russia and China are expected to deploy AI as a matter of routine when carrying out cyber-attacks.

The UK's critical national infrastructure takes in a broad set of sectors, spanning energy, transport, defence, food, health and finance. An attack on any of these areas could disrupt services that households and businesses depend on.

Ransomware has become a mainstay for criminal hacking groups, many of which are based in countries that were once part of the Soviet Union. The technique involves breaking into a victim's computer systems, scrambling files with custom-built malicious code and then demanding cryptocurrency in exchange for restoring access and returning any data taken. Previous UK victims have included Royal Mail, the outsourcing company Capita and a number of public bodies such as councils.

According to the NCSC, AI stands to sharpen the ransomware danger by assisting attackers with reconnaissance, finding gaps in networks and producing harmful code. The agency's chief executive, Richard Horne, is due to use a speech alongside the review to tell the public that a sharp increase in cyber incidents could arrive shortly. He is expected to link the risk to the technology's capacity to expose flaws that organisations have put up with for too long.

One illustration of that capability came last month, when AI helped uncover a serious vulnerability in PHP, a programming language behind close to 70% of websites worldwide. Alan Woodward, who teaches cybersecurity at Surrey University, said the warning showed why governments and firms need access to the most advanced AI tools, arguing that if legitimate defenders are shut out, criminals will get there first. Leading AI developers including OpenAI and Anthropic have already given some organisations early access to their strongest models to help spot serious IT faults.

Horne will also report that the communications networks relied on by the British military come under daily attack, describing the activity as persistent and sustained. He will name Russia among those responsible while noting the operations have been kept short of open warfare.