SUPPLY CHAIN MANAGEMENT
time to detect, contain and respond before the consequences spread to connected systems and partners.
Security & innovation
None of this means that speciality chemical companies should avoid AI, however. The potential benefits make adoption worth pursuing, but security and resilience must be built in from the beginning, rather than treated as obstacles to be addressed afterwards.
A sensible starting point is to understand where AI is being introduced and what it depends upon. Organisations should map the systems, data flows and third parties involved in an AI-enabled process and identify where it connects to business-critical or operational environments. The level of assurance should reflect the potential consequences of failure or compromise.
An AI tool with limited access to information and no ability to influence operational processes still provides significant benefits but presents a very different risk from one that can access sensitive production data, make procurement decisions or even control or interact with operational technology. The important questions are what the AI can access, what it can influence and what could
happen if it were to be compromised, became unavailable or produced an incorrect result.
Access should be tightly controlled, with AI systems given only the permissions they require. This is particularly important where they can interact with sensitive information or other systems. Interfaces between AI applications, cloud services, enterprise IT and OT environments also deserve particular attention, as they can create new pathways between systems that were previously more isolated.
Third party assurance is equally important. If a critical process depends on an external AI platform, the security and resilience of that provider become part of the organisation’ s own risk profile. Understanding those dependencies and what happens if a provider is compromised or unavailable should therefore form part of the wider supply chain assessment.
Finally, organisations need to plan for failure. For processes that matter to operational continuity, organisations need to know how they will continue to operate safely if the AI capability suddenly cannot be trusted. Resilience does not mean trying to ensure that AI will always work, but planning how to continue to operate when it does not.
System resilience
This shift in thinking is increasingly important as regulation and scrutiny around critical infrastructure and its supply chain develops. The direction of travel is towards greater attention to individual organisations and the dependencies that connect them. That makes supply-chain security an increasingly important component of operational resilience.
For speciality chemicals manufacturers, this is particularly relevant because their products can sit at the intersection of multiple industries and supply chains. It is neither realistic nor particularly useful to create an impenetrable boundary around every individual organisation in a highly connected industrial environment like this.
Instead, organisations need to understand how threats can move through the ecosystem, where the most consequential dependencies lie and where controls can prevent a local cyber incident from becoming a wider operational problem. AI makes this more urgent because it can increase the speed, scale and interconnectedness of digital activity, while also giving attackers new ways to automate existing techniques.
Speciality chemicals businesses need to look beyond the security of individual AI systems and consider what would happen to themselves and essential services that depend on them if the systems, or those around them, were compromised. As AI becomes embedded across chemical manufacturing and its supply chains, building resilience around those dependencies will be critical to ensuring that innovation strengthens rather than inadvertently weakens the systems that depend on them. ●
J j
Chris Anley
CHIEF SCIENTIST
NCC GROUP press @ nccgroup. com www. nccgroup. com
54 SPECIALITY CHEMICALS MAGAZINE ESTABLISHED 1981