SUPPLY CHAIN MANAGEMENT
An AI system could depend on information from an ERP platform, supplier databases, production systems or external cloud services, and it may also interact with other applications through APIs or be integrated into an operational workflow. Each of these connections creates a dependency and a potential point of attack.
That means manufacturers not only need to determine whether an individual AI application is secure in itself, but also understand what that application connects to, what information it can access, who operates the underlying technology and what happens if it becomes unavailable or compromised. This is particularly important where the organisation sits within a wider critical supply chain.
Extended supply chain
A chemical manufacturer rarely operates in isolation. It will depend on specialist software and technology providers, managed service providers, cloud platforms, rawmaterial suppliers, logistics partners and distributors, while its own products may form essential inputs for customers in other industries.
The result is a web of interconnected dependencies where a cyber incident affecting one organisation can disrupt others further along the supply chain. It is rare for organisations to fully understand the implications of all components of their supply chain. A real incident very often reveals unexpected dependencies.
If, for example, a manufacturer supplying a chemical essential to a water-treatment process faced a production or distribution outage following a cyber incident, the customer could face constraints on its ability to maintain normal treatment operations. In a worst-case scenario, this could affect the quality or availability of treated water and ultimately disrupt a service on which households, businesses and other essential services depend.
A crucial point here is that the supplier may not be aware that the product or service they supply forms this essential link in the supply chain. The same dynamic can apply across energy, healthcare, manufacturing, transport and electronics.
These dependencies can be easy to overlook when assessing resilience. A company may not operate critical infrastructure itself, but a disruption to its products or services could still affect organisations that do. Understanding those dependencies is therefore an important part of managing supply chain risk.
AI changes the equation
The introduction of AI can make these relationships more complex. A procurement team, for example, might use AI to analyse supplier information, forecast demand or identify purchasing opportunities. This could improve efficiency and reduce costs, but it also means that an AI-enabled process may influence decisions that were previously made through more traditional systems and controls. In manufacturing, AI may be introduced to analyse production data and identify anomalies that could indicate equipment problems. The benefits could include earlier intervention and improved operational efficiency.
However, if the AI system is connected to operational technology, its security and reliability become part of the wider operational risk picture. This is particularly important in chemical manufacturing, where new digital capabilities may sit alongside established operational technology and legacy systems that were not necessarily designed for today ' s highly interconnected environment.
There is also the question of third-party AI. Many organisations use external models, platforms and applications, potentially provided through a chain of technology suppliers. Each adds further dependencies in a growing and frequently updated software dependency tree.
This creates questions around data, access, security assurance, availability and the ability to understand how the technology behaves when something goes wrong. The risk is therefore not confined to the model itself; it extends across the chain of systems, suppliers, people and processes surrounding the model.
Attacker’ s playbook
There is another side to the AI equation. Threat actors can also use AI to make existing attacks faster, more scalable and more convincing. This could mean more sophisticated social engineering targeting employees, suppliers or customers.
Voice and video deepfakes can make social engineering attacks much more powerful. AI can help attackers analyse publicly available information and create highly tailored messages that appear to come from trusted individuals or organisations. In an industry where employees regularly interact with multiple external partners, distinguishing legitimate communications from malicious ones can become increasingly difficult.
AI can also help attackers to scale activities that previously required more time and effort. Reconnaissance, gathering information and preparing targeted attacks can all be accelerated, increasing the pressure on organisations that are already managing complex technology estates and large networks of suppliers.
The implications extend beyond the manufacturer itself. An attacker looking to disrupt a critical supply chain does not necessarily need to compromise its ultimate target directly, as a smaller supplier may offer another, quieter, less well-defended route into the wider ecosystem.
This is where the systemic nature of the risk becomes particularly important. If AI allows an attacker to identify and exploit weaknesses more quickly, organisations have less
NOV / DEC 2026 SPECCHEMONLINE. COM
53