Collective push for AI standards

Extensive work is currently underway on the standardisation of artificial intelligence (AI), nationally and internationally. Standards are necessary to be able to deal with the opportunities and challenges associated with the use of AI.

"Standards create predictability, demonstrate that elements work together, and by no means least, standards contribute to safety," says Linn Iren Vestly Bergh, who works specifically on digitalisation and new technologies at Havtil.  

In recent years, she has dedicated much of her time to AI - primarily to how the authorities can contribute to promoting the prudent development and use of AI in the industry. 

Vestly Bergh participates actively in standardization work both nationally and internationally. This involves attending meetings, providing input and assessing draft standards. In addition, she is engaged in sharing knowledge about standardisation internally at Havtil and out to the industry. 

"Knowledge of the work on standardisation is important for us as a supervisory authority to be able to follow up AI in the industry in a positive manner. This includes having up-to-date regulations and supervisory methods ensuring that the industry uses AI in a prudent manner.”

Standards create predictability, demonstrate that elements work together, and by no means least, standards contribute to safety.

Sorting jobs

Vestly Bergh admits that it can be difficult to remain silent on the matter of what is what within this work scope. By way of example, how does the EU's AI regulation relate to standardisation work in practice? 

“The AI Regulation, which the EU is working on, sets the overall framework for how AI should be developed and used. It defines risk levels and requirements for systems that may pose a high risk to society.

Standardisation is an important tool in terms of meeting the regulatory requirements," she comments. 

The EU regulation thus states what must be achieved, whilst standards can provide technical specifications for how to achieve it. The standards can, for example, describe important principles and procedures for testing AI systems, assessing datasets, or ensuring transparency and human-centric design.

Standards can also help to harmonize practices across national borders, which will become important in a globalized industry. 

Clear call

Vestly Bergh is Havtil's representative in CEN-CENELEC Joint Technical Committee 21 (JTC 21), a standardization committee which aims to develop harmonised standards for AI technologies. 

The work of JTC 21 is divided into several working groups, which target specific areas. These include, for example, AI terminology, risk management, data management, cybersecurity, and human supervision and control. Companies which choose to comply with the harmonised standards will, as a general rule, be able to assume that they comply with the AI Regulation.  

“The development of standards is basically a consensus-based process, which means that all parties must be in agreement before a standard can be adopted. It's a lot of work, but it's also very rewarding to be a part of shaping the standards of the future," says Vestly Bergh, who has a clear message for the industry: 

“Become involved in this work. AI technology is developing at speed, and we must work together to ensure that we have the necessary standards in place to keep up. Without industry standards, we risk a fragmented approach to AI security. This may lead to the development and implementation of AI systems which increase the risk of accidents and incidents. 

It is also important to build trust in AI technology. Clear and unambiguous standards fostering optimal AI solutions will also promote trust,” Vestly Bergh points out.  

 “The sooner we put standards in place, the better equipped we will be to handle the challenges and opportunities provided by AI.”

Standard

A standard can be defined as a "document for common and repeated use, developed by consensus and approved by a recognized body, which provides rules, guidelines, or characteristics for activities or their results, aimed at achieving optimal order in a given context." (Source: Standards Norway)

Harmonised standard

A harmonised standard is a standard developed by a recognised European standardisation organisation on behalf of the European Commission.  

If the requirements of a harmonised standard are met, it can be assumed that the relevant requirements of EU legislation are fulfilled. 

The EU determines whether a standard is harmonised.  

European AI standardisation

CEN-CENELEC Joint Technical Committee 21 (JTC 21) is a joint technical committee for the standardization of artificial intelligence. 

The committee was established in 2021 by the European standardisation organisations CEN (Comité Européen de Normalisation) and CENELEC (Comité Européen de Normalisation Électrotechnique). 

JTC 21 has a mandate from the European Commission, and the standards are intended to help create a common understanding of and a uniform approach towards AI regulation in Europe.  

The work in JTC 21 supports the implementation of the EU AI Regulation by developing harmonised standards.