Public trust in AI systems has been seriously affected by a number of scandals where the use of AI led to biased, harmful and inaccurate decision-making. Civil society organisations and regulators thus demand a way to ensure the trustworthiness of AI systems. Guarantees are needed (e.g. from auditors) that a system is trustworthy, so among others reliable and fair, if it is to be employed in high risk settings. This project aims to enable that trustworthiness assessments of AI systems. To achieve that goal our objectives are twofold. First, to develop a robust, theoretically sound pipeline for quantifying the risk associated with AI systems in both decision-support and generative AI settings. Second, to collaboratively develop trustworthiness assessment standards with civil society organisations in order to identify their needs and ensure that the pipeline empowers them to check the use of AI. These two objectives will be carried out in an iterative feedback loop, designed to align the prototype pipeline with co-created societal requirements
