Ethical Charter for Artificial Intelligence

Our commitment to trustworthy AI

Artificial intelligence is a technology that is bringing about a profound transformation of organisations, professions, services and digital uses.

OLING supports its clients in their transformation, information systems and data protection projects. In this context, the use of artificial intelligence solutions is essential and must form part of a responsible and controlled approach that respects individuals and our clients’ data.

This Ethical Charter for Artificial Intelligence defines the principles that OLING undertakes to apply when using, assessing, designing, integrating or recommending solutions based on artificial intelligence. It applies both to internal uses of AI and to services provided to our clients.

It supplements OLING’s policies and procedures relating in particular to information security, personal data protection, confidentiality and risk management.

1. AI with a defined purpose and controlled impact

Any use of artificial intelligence must serve a clearly identified, legitimate and proportionate purpose.

Before using or recommending an AI solution, OLING takes care to identify its objective, the users concerned, the expected results and the main risks that may arise from it.

AI must be used in the interests of individuals and organisations and must not create or amplify unjustified infringements of rights and freedoms.

Where justified by the nature or sensitivity of a project, the potential impacts of AI are assessed before its implementation and appropriate measures are defined to control the identified risks.

2. Responsible and sustainable AI

OLING takes into account the environmental, social and organisational impacts associated with artificial intelligence solutions.

The use of AI must be proportionate to the need and provide genuine value in view of the resources required for its operation.

Where several solutions can meet the same need, OLING encourages consideration of criteria such as the location of solutions, resource consumption, the long-term viability of the solution, its usefulness and its impacts on stakeholders.

The pursuit of technological performance must not be separated from the pursuit of responsible and sustainable digital use.

3. Fair and non-discriminatory AI

OLING takes care to ensure that the use of artificial intelligence does not lead to unjustified discrimination or unfair treatment.

Particular attention is paid to the quality and representativeness of data, as well as to biases that may be introduced by the data, models, operating rules or conditions of use.

Where results produced by AI may have a significant impact on individuals, the risks of bias and discrimination must be identified and, where possible, assessed and corrected.

Diversity of skills and perspectives is encouraged in the analysis and management of projects using artificial intelligence.

4. Transparent and explainable AI

OLING promotes the transparent use of artificial intelligence.

Where a person interacts directly with an AI system or where content is produced to a significant extent by AI, they must be informed of this use where the context requires it.

The capabilities and limitations of the solution must be understood by the people who use it.

Decisions and results produced with the assistance of AI must, where warranted by their importance, be explainable, traceable and subject to review. The expected level of explainability is adapted to the nature of the use and the seriousness of the potential consequences of an incorrect result.

5. AI under human oversight and with clearly defined responsibilities

Artificial intelligence is an assistance tool and does not transfer responsibility to the machine.

People using an AI solution remain responsible for how they use it in the performance of their duties. Appropriate human oversight must be maintained where results produced by AI may have significant consequences.

Results generated by AI must not be assumed to be accurate. They must undergo a level of verification appropriate to their use.

The roles and responsibilities of the various parties involved must be identified, particularly where several parties participate in the provision, integration or operation of an AI solution.

6. Robust and secure AI

Information security is taken into account when selecting, using and integrating artificial intelligence solutions. OLING applies security measures to AI systems that are proportionate to the risks associated with their use.

Particular attention is paid to the confidentiality of information, access control, data integrity and quality, reliability of results, vulnerability management, dependency on suppliers and business continuity.

Sensitive, confidential or client-owned information must not be entered into an AI solution without first ensuring that its use is authorised and that the security safeguards are appropriate.

Where necessary, it must remain possible to revert to human or conventional processing.

7. AI that respects privacy and personal data

OLING applies the principles of personal data protection and privacy by design to uses of artificial intelligence.

Any processing of personal data carried out using an AI solution must comply with the applicable data protection requirements.

In particular, OLING takes care to limit the data used to what is necessary for the intended purpose, to ensure its security and to determine the conditions under which it may be disclosed to an AI solution provider.

Particular vigilance is applied to sensitive data, confidential information and processing operations likely to present significant risks to individuals. Data subjects must receive the appropriate level of information and be able to exercise their rights under the conditions laid down by the applicable regulations.

Implementation of the Charter

These principles apply to OLING employees and to parties acting on its behalf when they use or participate in a project involving an artificial intelligence solution.

AI solutions and suppliers used by OLING must be selected with due consideration for security, confidentiality and data protection requirements and the principles defined in this Charter.

As part of the assignments carried out for its clients, OLING takes care to identify the respective responsibilities of the various parties and to alert the client where an intended use of AI presents an identified significant risk.

Any employee who has doubts concerning the use of an artificial intelligence solution must seek approval before continuing to use it.

This Charter is communicated to the persons concerned and is reviewed periodically, particularly to take into account developments in uses, technologies, risks and the regulatory framework.