Towards Trustworthy AI in Europe
The objective of the EU is to secure the uptake of algorithmic systems in Europe, especially those based on artificial intelligence (AI), while ensuring the protection of the safety and fundamental rights of European citizens.
AI made in Europe should be trustworthy: transparent, fair and human-centred, with a positive societal impact.
Understanding the social impact of AI
In this research topic, the JRC:
- advances the scientific understanding of machine and human intelligence,
- studies the societal and ethical impact of algorithms,
- defines methodologies for trustworthy AI,
- and provides scientific contributions to EU AI policies.
This work addresses different types of AI systems, such as recommender systems, facial processing systems, conversational systems or automated driving systems. The JRC maps the opportunities these systems offer to different application domains, assesses risks that these systems bring to safety, fundamental rights and mental well-being, develops practical methodologies for trustworthy AI, evaluating and ensuring transparency, fairness and human oversight of those systems and studies the long-term impact of algorithms, such as the impact on children or mental health.
Latest developments
A recent JRC paper explores AI benchmarks, considered an essential tool to evaluate performance, capabilities, and risks of AI models. Through a comprehensive literature review, the paper identifies key shortcomings of AI benchmarking, as well as policy approaches that could mitigate these.
The AI Office invites experts to express their interest to participate in the drawing-up of the Code of Practice.
New study features over 140,000 annotations of protected attributes in widely used visual datasets for perception, and a thorough evaluation of biases and the annotation process.
Related policy initiatives
This topic allows the JRC a direct contribution to two core policy initiatives in the different stages of the policy making process:
-
The AI Act
-
The Digital Services Act (in the context of the European Centre for Algorithmic Transparency).
Our work also supports AI policies in specific sectors such as transport, education or culture.