More than a year has passed since the Regulation (EU) 2024/1689 (known as the AI Act) came into force, establishing harmonized European rules on artificial intelligence, and some of its provisions are already mandatory.
For those involved in legal metrology, some questions arise spontaneously: is the world of legal metrology (measuring instruments, notified bodies, testing, market surveillance, etc.) currently being considered in this regulation? If so, how or in which aspects? Or are these two areas that do not interact with each other?
This article attempts to offer an initial response to these questions, enriching it with some insights and unexpected observations. Of course, given that the matter of the AI Act is extremely broad, the article does not intend to and cannot cover all aspects, but it aims to introduce the topic and then develop it in subsequent articles.
AI Act and New Legislative Framework
A very useful introduction to understanding the subject is to consider that the AI Act is fully integrated into the New Legislative Framework, the new European legislative framework for a new approach [1], adopted with Regulation (EC) 765/2008, which harmonizes European legislation on products placed on the EU market, accreditation, and market surveillance. The European directives concerning legal metrology measuring instruments (NAWID 2014/31/EU, MID 2014/32/EU) also belong to this same approach.
Consequently, the AI Act, in addition to complying with this approach and incorporating all its principles, does not prejudice or in any way hinder the rules laid down by other European Union legal acts on consumer protection, product safety, and other fundamental rights. On the contrary, the AI Act completes the legislative framework of the new approach [2].
More clearly, the same principles and mindset that underpin the Directives NAWID 2014/31/EU and MID 2014/32/EU also inspire the AI Act.
In short, what is the AI Act and what is its purpose?
The AI Act is a European Union regulation that is directly applicable in each Member State (without the need for transposition), establishing a uniform legal framework in Europe for the development, placing on the market, putting into service, and use of human-centric and trustworthy artificial intelligence (AI) systems and related AI models.
In practice, it establishes the requirements that these AI systems must meet in order to enter the European market, ensuring a high level of protection of health, safety, and rights enshrined in the Charter of Fundamental Rights of the European Union; how to verify their conformity, who can verify it, based on which technical specifications/standards; how to monitor AI systems after they enter the market and how to ensure their safe technological development.
The object of the regulation is AI systems and general-purpose AI models [3] considered both as a complete product in themselves and as a component of another product that is placed on the market or put into service. AI systems are classified according to their potential impact on safety, fundamental rights, and the values of the European Union.
For high-risk AI systems [4], conformity assessment (with or without the involvement of a notified body), CE marking, declaration of conformity, and registration (in an EU database or at national level) are required.
For general-purpose AI models and those with systemic risk [5], conformity assessment does not involve notified bodies.
The current classification of high-risk AI systems and general-purpose AI models with systemic risk may be always updated and modified by the European Commission following new assessments or conditions, as well as technological progress.
The entity (economic operator) primarily responsible for placing AI systems and models on the market/putting them into service is the “provider,” means a natural or legal person, public authority, agency or other body that develops or has developed an AI system or a general-purpose AI model and places it on the market or puts the AI system into service under its own name or trademark, whether for payment or free of charge [6].
The conformity of AI systems with the AI Act:
a) is presumed if harmonized standards or common specifications and codes of practice (the latter in the absence of harmonized standards) are observed;
b) is assessed, for high-risk AI systems, through an internal control carried out by the provider, or with the involvement of a notified body based on the assessment of the quality management system and technical documentation.
Implications for manufacturers of legal metrology measuring instruments
In practice, how are the manufacturers of legal metrology measuring instruments, referred to in the European directives NAWID 2014/31/EU and MID 2014/32/EU, affected by the provisions of the AI Act? Two aspects should be highlighted immediately.
| 1. Legal metrology measuring instruments incorporating AI systems |
| Currently, the AI Act, in classifying high-risk AI systems, takes into account both those that may be used as a safety component of a product regulated by a specific list of EU harmonization legislation (directives and regulations) or subject to third-party conformity assessment prior to being placed on the market, as well as those used in specifically listed areas (AI Act, Annex III). Well, although consumer protection, which inspires the legal metrology directives, is undoubtedly a fundamental value for the European Union (as is security in commercial transactions), the NAWID 2014/31/EU and MID 2014/32/EU directives are not included in the list of harmonized regulations that determine the classification of high-risk AI systems, which could be integrated into legal metrology instruments. To date, the only case covered by the AI Act in which an AI system is considered high-risk if it is a safety component of a legal metrology instrument is that of the ‘critical infrastructure’ area, means ‘ AI systems intended to be used as safety components in the management and operation of critical digital infrastructure, road traffic, or in the supply of water, gas, heating or electricity’ [7]. Therefore, in the case highlighted, this certainly includes all smart meters (regulated by Directive MID 2014/32/EU) with integrated AI systems. While AI systems used as a safety component or otherwise in all other legal metrology instruments are not currently classified as “high-risk.” Consider those that can integrate fuel dispensers (petrol, LPG, CNG, LNG, hydrogen, electric), automatic weighing instruments, measurement and payment software, etc. If the legal metrology instrument has a high-risk AI system as a safety component, the manufacturer of that instrument assumes the more specific responsibilities and obligations of the deployer or even the AI system provider (see point 2) [8]. In general, AI systems connected to products that are not high-risk under the AI Act, and therefore not required to comply with the requirements established for high-risk AI systems, it is important that are nevertheless safe when placed on the market or put into service [9]. Therefore, whenever a legal metrology measuring instrument uses AI systems, it is advisable for the manufacturer to assess the extent of its responsibilities and the actions necessary to comply with the AI Act. In summary. ➡️ If a non-high-risk AI system is used in the measuring instrument, as a deployer or provider (see point 2), the manufacturer is required to carefully assess the extent of its responsibilities and the actions necessary to comply with the AI Act, as well as to monitor any changes to the classification rules of the AI Act itself, which could cause the AI system used to fall into the “high-risk” category, with the related consequences. ➡️ If a high-risk AI system is used in the measuring instrument as a safety component of an instrument used in the ‘critical infrastructure’ area, the manufacturer must assess its position as a deployer or provider and, only if it falls into the latter category, also carry out the conformity assessment of the AI system (following the assessment procedure based on internal control without the involvement of a notified body), the EU declaration of conformity, the EU marking, and the registration at the national level. |
| 2. Manufacturer of the measuring instrument, deployer and provider of AI systems |
| What is the connection between the role of the manufacturer of legal metrology measuring instruments and the roles typically defined in the AI Act for deployers and providers of AI systems? 🟣 The manufacturer of the measuring instrument, as is well known, is any natural or legal person who manufactures or has designed or manufactured a measuring instrument, and markets that measuring instrument under his name or trade mark or puts it into use for his own purposes. 🟣 The deployer is any natural or legal person, public authority, agency or other body using an AI system under its authority in the course of a professional activity [10]. 🟣 The provider, under the AI Act, is any natural or legal person, public authority, agency or other body that develops an AI system or a general-purpose AI model or that has an AI system or a general-purpose AI model developed and places it on the market or puts the AI system into service under its own name or trademark, whether for payment or free of charge [11]. Although these three entities are quite different, they can overlap and sometimes even merge. When does this happen? First, the AI Act applies if the manufacturer of legal metrology measuring instruments places an AI system on the market or puts it into service together with its product and under its name or trademark [12]. The manufacturer of measuring instruments is considered to be a provider of AI systems: a) if it develops or has developed an AI system and places it on the market/in service under its own name or trademark; b) if it puts its name or trademark to a high-risk AI system already placed on the market/in service, or makes a substantial change to it such that it remains a high-risk AI system; c) if it modifies the intended purpose of an AI system not classified as high-risk and already placed on the market/in service in such a way that it becomes a high-risk AI system. The manufacturer of the legal metrology measuring instrument assumes the role of deployer when using an AI system in the context of its professional activity, with particular attention to the responsibilities involved in the case of a high-risk AI system. In summary, as per the New Legislative Framework approach, the criteria for qualifying as a manufacturer of a measuring instrument or a provider of an AI system are always based on the same ratio: the affixing of one’s name or trademark and the substantial modification of the product (measuring instrument or AI system). |
Legal metrology in support of the technical specifications of the AI Act
An interesting, and perhaps unexpected, link between the AI Act and the legal metrology is the support that the latter is called upon to provide in defining the technical specifications used to define the requirements of AI systems in order to comply with the AI Act. In what sense?
As is well known, one of the key requirements on which legal metrology measuring instruments are designed and then verified is the measurement accuracy.
Well, one of the requirements that the AI systems (especially high-risk ones) must meet in order to be compliant is an appropriate level of accuracy [13].
Since the technical specifications for proving the compliance of AI systems with the requirements of the AI Act are not yet fully defined, the European Commission not only expressly involves legal metrology, but also encourages its assistance in developing benchmarks and measurement methodologies for AI systems.
The recital (74) and article 15 of the AI Act are very explicit: ‘Union law on legal metrology, including Directives 2014/31/EU and 2014/32/EU of the European Parliament and of the Council, aims to ensure the accuracy of measurements and to help the transparency and fairness of commercial transactions. In that context, in cooperation with relevant stakeholders and organisation, such as metrology and benchmarking authorities, the Commission shell encourage, as appropriate, the development of benchmarks and measurement methodologies for AI systems. In doing so, the Commission should take note and collaborate with international partners working on metrology and relevant measurement indicators relating to AI’.
In fact, European and international legal metrology groups are already working to provide significant support for the development of technical specifications for AI systemsregarding accuracy.
And so, the legal metrology sector, which is often slow to keep pace with regulatory innovation, is now rapidly becoming involved in one of the most innovative sectors of our time: artificial intelligence.
Insights for future developments
What could be the future developments involving legal metrology, artificial intelligence, and the AI Act? Here are some insights.
| An update of the list and the classification rules for high-risk AI systems could, in the future, include legal metrology measuring instruments for three main reasons: 1) greater consideration of consumer protection, which is a fundamental value for the European Union, the basis on which the legal metrology directives are inspired, and repeatedly mentioned in the AI Act, so much so that it is established that it should never be prejudiced by its own harmonized standards [14]; 2) the observation that measuring instruments (their components and software) in all areas of legal metrology increasingly use and will use AI systems; 3) the fact that legal metrology measuring instruments are subject to third-party conformity assessment for placing on the market or putting into service in accordance with specific European Union harmonization regulations (NAWID and MID directives). This reason is precisely one of the rules for classifying high-risk AI systems already covered by the AI Act (art. 6, paragraph 1, letter b). |
| The AI systems could certainly be used for the conformity assessment of legal metrology measuring instruments in accordance with Directives NAWID 2014/31/EU and MID 2014/32/EU, both by the notified body and by the manufacturer. The notified body could use AI systems to perform type conformity tests (known as module B), while the manufacturer could use them to manage the quality of the production process (known as module D). So, what is the impact of AI systems in this context? Should they be taken into consideration for the purposes of the AI Act? |
| Last but not least, market surveillance authorities could also use AI systems in various ways for their post entry into market checks (periodic verification and random checks) of legal metrology measuring instruments. What would be the consequences, given that market surveillance authorities ensure that measuring instruments remain compliant, protecting consumers and safeguarding public interests [15]? How should AI systems used in market surveillance activities be evaluated? |
Conclusions
In conclusion, and returning to the initial questions, it can be said that there are at least three aspects that link the legal metrology and the AI Act 2024/1689.
The first: the contribution that legal metrology is called upon to make in identifying methodologies for measuring the accuracy of AI systems.
The second: the fact that measuring instruments, and especially the software within them, use AI systems, and therefore may be subject to the AI Act, involving the manufacturers of measuring instruments at various levels of responsibility.
The third: the use of AI systems in procedures for assessing compliance with the NAWID and MID directives, as well as in market surveillance activities. The latter aspect does not appear to be considered in the AI Act to date, although it is relevant for the protection of public interests.
Of course, the application of the AI Act has only just begun and much remains to be tested in many respects. Only practice will show us the way to achieve the right balance between regulatory requirements, protection of the interests of all stakeholders (economic operators, consumers, authorities), and technological progress. It is up to each of us to become a protagonist in this revolution and contribute to the reliability of AI, including in the world of legal metrology.
Notes
[1] For an overview of the principles, concepts, and applications of the new approach of the European legislation, see the latest edition of the “Blue Guide” on the implementation of EU product legislation.
[2] Regulation (EU) 2024/1689 (AI Act), recital (9); art. 2, point 9.
[3] For the definitions of ‘AI system’ and ‘general-purpose AI model’, see art. 3, points 1) and 63), and recital (97) of Regulation (EU) 2024/1689 (AI Act).
[4] The classification of high-risk AI systems is in art. 6, Regulation (EU) 2024/1689 (AI Act).
[5] For the definition of ‘systemic risk’, see art. 3, point 65), Regulation (EU) 2024/1689 (AI Act).
[6] Regulation (EU) 2024/1689 (AI Act), art. 3, point 3).
[7] For the classification of high-risk AI systems, see Regulation (EU) 2024/1689 (AI Act), art. 6, Annex I and Annex III (in particular point 2 for the legal metrology).
[8] Regulation (EU) 2024/1689 (AI Act), artt. 25, 26.
[9] Regulation (EU) 2024/1689 (AI Act), recital (166).
[10] Regulation (EU) 2024/1689 (AI Act), art. 3, point 4).
[11] Regulation (EU) 2024/1689 (AI Act), art. 3, point 3).
[12] Regulation (EU) 2024/1689 (AI Act), art. 2, letter e).
[13] Regulation (EU) 2024/1689 (AI Act), artt. 13 and 15.
[14] Regulation (EU) 2024/1689 (AI Act), recitals (9), (45), (48), art. 2 par. 9.
[15] Regulation (EU) 2019/1020 on market surveillance and compliance of products.
Reference source
European Union web site, Regulation (EU) 2024/1689 (AI Act).
