Back

Certification of processes for AI

In response to the growing number of artificial intelligence solutions, LNE has created a certification that provides users with objective criteria for making their selection, and enables developers to demonstrate that they have mastered all the stages of the AI life cycle and meet the performance, regulatory, confidentiality and ethical requirements of their customers.

Responding to issues of trust

Artificial intelligence (AI) has developed strongly in recent years in many professional sectors (collaborative industrial robots, inspection and maintenance robots, autonomous mobility systems, etc.) and in the home (personal assistance robots, medical devices, personal assistants, etc.), making it necessary to be able to reliably demonstrate the levels of performance, robustness, ethics and explicability achieved by the different AI solutions.

Thanks to our dual expertise in certification and evaluation of AI algorithms, LNE has decided to create a voluntary certification in this field that meets the needs of market players.

Market players	Requirements	AI certification responses
End users and buyers of AI solutions	Having confidence in the performance of the systems they acquire The ability to choose from a multitude of existing solutions	A common, objective and clear reference to make their choice Guarantees allowing to increase the acceptability of these technologies
AI developers, suppliers, integrators	Stand out from competitors Guarantee the performance of their algorithms or systems	Demonstrate that they master all the stages of the AI life cycle and meet the requirements defined by their clients in terms of performance, confidentiality, ethics, as well as regulatory requirements Use as a benchmark to guide their R&D and control efforts

Certification, the outcome of a consensus

To build the certification standard and meet the need for trusted solutions, LNE has set up (in 2020) a working group representative of the AI community, bringing together developers, evaluators and end users of algorithms. It also collected opinions through a public call for comments.

Click for more information

A certification of processes

This standard aims to define common requirements for the design, development, evaluation and maintenance processes of all types of AI functionalities using machine learning.

It therefore covers all business sectors in which AI is used, in order to ensure the application of best practices that promote confidence in AI. It is not intended to define requirements for AI functionalities and therefore specific to their uses.

The design, development, evaluation and maintenance processes covered by the standard are defined as follows:

Certified procedures	Definition	Examples of audited procedures
Design	Transforming an expression of requirements into functional specifications	Specification and inclusion of normative and regulatory requirements
Development	Translate these specifications into a version of the AI functionality ready for evaluation	Learning Quality of Databases
Evaluation	Verify the conformity of the system to the specifications defined before its deployment	Definition of evaluation protocols, metrics, in every evaluation tools to allow reporting on the effectiveness of these intelligent systems
Maintaining operational conditions	Ensure compliance of the AI functionality with the defined specifications after its deployment and throughout its operating phase	All features necessary to maintain operating conditions AI systems can evolve throughout their life with performance degradations

Why opt for certification?

Developers, suppliers and integrators of AI solutions, thanks to certification, will benefit from multiple advantages allowing them to bring an optimum guarantee of confidence to their customers.

Demonstrate that they master all the stages of the AI life cycle and meet customer requirements for performance, regulatory compliance, confidentiality and ethics
Attest to the quality of development of their products containing AI
Establish a high level of confidence in their solutions by mastering the processes and best practices of the AI domain
Demonstrate that they have an organisation, structure and processes in place to produce quality and trusted AI over the long term
Have a competitive advantage
Benefit from optimal credibility by being audited by an independent and competent body, thus ensuring a good level of confidence in the respect of the requirements

Obtaining the AI certification

To obtain AI certification, developers, suppliers and integrators of AI solutions must meet the requirements of the "Certification standard of processes for AI: design, development, evaluation and maintenance in operational conditions".

The standard

The certification standard of processes for artificial intelligence defines:

The certification procedure, including the scope of application
The requirements to be met by the applicant/owner
Admission, supervision and staffing arrangements

Procurement of the certification is governed by § 2 of the standard.

Download the AI certification standard

The steps to AI certification

AI certification audit cycle

After obtaining your certification, you enter in a cycle of annual audits:

An initial follow-up audit to be carried out at 12 months
A second follow-up audit to be carried out at 24 months,
A renewal audit after 3 years

Applying for certification

For any request with LNE Digital certification, artificial intelligence, contact us by clicking on "Obtain a quote".

LNE's strong points

Over 10 years of experience in evaluating intelligent technologies
More than 900 evaluations of artificial intelligence systems, including language processing, image processing and robotics
Expert in the constitution and use of large, quality, structured and annotated test databases
Participation in the AFNOR commission on artificial intelligence, in the AFNOR strategic committee on information and digital communication and in section 81 of the UNM on industrial robotics
Participation in the development of key resources for AI assessment and development

LNE clients testify

Certifier notre solution Infracare nous a permis de rassurer nos clients potentiels et de leur démontrer notre niveau d’expertise...

Sébastien Denaës

Chef du service Équipements, Méthodes et Data Science au sein du Centre de recherche de Colas

Témoignage complet

En plus d’être une étape essentielle de notre préparation à l’AI Act, notre méthode projet certifiée nous permet de créer des solutions IA dignes de confiance.

Matthieu Capron et Aymen Shabou

Responsable de la Design Authority IA et CTO du DataLab Groupe Crédit Agricole