Tributech: AI is rewriting the threat model for Europe's CNI

Much of Europe's critical national infrastructure (CNI) was designed before today's AI-enabled, remotely operated threat landscape emerged.

Operators of CNI face a threat landscape their existing monitoring tools weren't designed for, argues Thomas Plank, CEO of London-headquartered cybersecurity firm Tributech. AI is reducing the resources needed for sophisticated attacks.

Plank's company works with operators in oil and gas, energy, utilities, manufacturing and pharmaceuticals that typically run Internet of Things (IoT) and operational technology (OT) assets, systems that run physical processes critical to their business.

"Most of these companies have the traditional cybersecurity tools in place mainly monitoring tools to get visibility within their environment, which was a good approach until a couple of years ago," he said. "Now with a changing threat landscape and all of the modernisation happening in CNI, there are new attack vectors which can't really be covered in full by the existing approach."

In CERT-EU's Threat Landscape Report for 2025, the organisation reported at least 174 threat actors engaged in malicious activity against the European Union ecosystem, with firewalls, VPNs, and network management solutions as primary targets.

The UK's National Cyber Security Centre handled 204 nationally significant incidents in the 12 months up to August 2025, up from 89 the year before.

Under the UK's Cyber Security and Resilience Bill, which updates the country's 2018 network and information systems rules, essential services will be required to report significant incidents within 24 hours and submit a full report within 72 hours. Plank said this puts significant pressure on companies to establish the facts and work out what happened before reporting. The House of Lords will consider the Bill on 26 October 2026.

The government's factsheet describes the initial notification as light-touch, covering the organisation's name, the affected service, and brief details, while informing the National Cyber Security Centre alongside the regulator. 

While the previous regime did not cover managed service providers, Plank said the rules will apply not only to the companies themselves but also to the providers they depend on. Essential services such as energy, transport, water, health and digital infrastructure will be included.

Plank said the risk comes from state-sponsored attackers, smaller groups and individuals. Many CNI operators map their risks to the four levels of the IEC 62443 industrial cybersecurity standard, with level four representing a state-sponsored attacker, he said.

"If we assume a level four is a state-sponsored attacker, high budgets, a lot of resources, high expertise, this can now be done by a smaller group of people or even individuals with AI tools, and the state-sponsored attackers also now have AI tools. It's an amplification of the damage that can be caused with little to no amount of resources," said Plank.

Plank described Zero Trust, a security model in which nothing is trusted by default, as the way to address this growing threat. In his view, each data point and command should not be trusted unless it has been verified, rather than being accepted because it comes from a trusted source or network. He said the question is whether it is genuine and whether it is allowed to be executed.

In a statement, Tributech said it extends zero trust from the network to the data layer: data is signed at the source and verified end to end, and commands are verified before they reach the asset.

Plank said building AI in-house can reduce risk because the operator owns the process from data collection through training to deployment. Buying off the shelf or using a partner makes the operator dependent on how the supplier developed the model or application. That being said, a specialist supplier might have stronger security practices, he said, but the operator doesn't own the process and has less traceability into what happened.

Either way, he said, operators need to verify the data going into AI and the commands coming out of it. Adding, operators should start at the boundary between OT and IT, one of the most vulnerable and most attacked areas, because an attack that starts in IT can move into OT, where the damage can be significant.

The past air-gapped approach, which keeps control systems physically isolated, may not suit modernisation initiatives such as real-time data visibility and over-the-air security patches. Over time, operators should move closer to the data source to add trust and reduce open attack vectors, Plank added.

"If something goes wrong in a CNI company, there might even be human lives on the line. In a traditional enterprise, if there is an attack, there might be only financial downsides or financial damage, but not really impact on people's lives, environment or even larger societies."