Ethical creation and use of AI technologies within Givaudan

Givaudan’s principles

The concept of Responsible AI (as defined below) encompasses the development and procurement of AI systems that are not only technically robust but also ethical, transparent, and understandable. It emphasises the integration of trust and equitable outcomes into the development of AI capabilities. It also encompasses the use of AI systems by Givaudan employees.

In a specific AI system, not all these principles can be implemented at maximum. For instance, there might be a compromise between privacy and transparency. The weighting of the principles varies depending on the purpose of the system, its users, and the context.

The business and IT owners of a specific AI system are responsible for making this compromise. For instance, a certain group of users might favour a principle over another of the list. These compromises must be documented and, if there is any doubt, they must be discussed with the AI Governance Board for validation (this weighting principle is inspired by the Data Science Swiss Federal guidelines).

With the use of ‘must’, the following sections define the minimum requirements for each principle, whereas ’should’ defines the maximum.

Givaudan will assess the risk linked to each AI system. This assessment will respect the risk classifications of the local regulations where our AI systems are used and operated.

AI systems assessed ‘high risk’ (whether bought or made), because of their potential impact, may have specific extra constraints listed in the following sections.