Published 25/02/26
Large Language Models (LLMs) such as ChatGPT, along with other AI-generated answers now commonly provided by search engines, are increasingly used by businesses to obtain fast guidance on regulatory topics. While these tools can be helpful for gaining a general understanding, relying on them for advice on the legal requirements of CE marking carries significant risks.
CE marking is governed by a complex and evolving framework of European regulations and directives, including the Machinery Directive, Low Voltage Directive, EMC Directive, Medical Devices Regulation, and many others. Anyone who has spent time looking through this legislation will recognise that the language used and nuance given within the legislation is crucial to understanding the legal responsibility of those looking to CE mark their products. This also applies in the context of UKCA marking products under UK regulations.
Whilst AI tools may be able to reference the correct legislation, they will also happily reference any document that it considers related. This will include national implementations of directives, superseded versions of the legislation, any documents that are published during the creation of the legislation and in the worst case, other irrelevant but similarly worded texts. In doing so it can return an answer, which sounds exactly like the kind of text you would find in the legislation, whilst completely missing the mark and removing (or worse adding) the nuance surrounding the requirements from its answer. In a regulatory environment where compliance depends on precise legal interpretation, even small errors can lead to serious consequences.
A further challenge lies in the lack of product-specific context. CE marking obligations depend heavily on the type of product, its intended use, operating environment, and the markets in which it is placed. AI tools typically provide generic guidance that cannot account for these variables. This can lead to incorrect assumptions, such as believing that certain directives do not apply, that self-certification is always allowed, or that testing requirements are minimal. In reality, many products require involvement from notified bodies, bespoke risk assessments, and extensive technical documentation. Misinterpretation at this initial stage can result in non-compliant products, enforcement action, product withdrawals, or costly redesigns.
There are many examples of specific requirements or exceptions across the CE marking legislation that are either product or legislation specific. Our customers rely on us to navigate and pinpoint the specific requirements for their products on topics such as Partly Complete Machinery, predominantly electrical equipment which could be considered machinery, updated and modified products, and the differences between CE and UKCA marking. While AI tools can assist with background research and conceptual understanding, they cannot replace expert interpretation or technical assessment.
For these reasons we should not rely upon Large Language Models or AI-generated search results as definitive sources of CE and UKCA marking advice. Robust compliance depends on accurate interpretation, technical expertise, and structured conformity assessment, all of which remain beyond the capabilities of AI systems, for now at least.
Please contact us for further information or for support.