In this ELOQUENCE webinar session, the discussion focused on one of the most complex and sensitive questions in artificial intelligence today: how can AI systems be developed in a way that respects human rights, ethical principles and European values?
The session featured Professor Martin Scheinin from the European University Institute, whose work in the ELOQUENCE project focuses on assessing whether emerging AI outputs comply with human rights, ethics and broader European values. With a long background in international law and human rights, he brought a legal and societal perspective to questions that are often discussed mainly from a technical point of view.
The first part of the webinar placed AI within a broader global context. Martin reflected on the role of digital technologies and AI in today’s “polycrisis”, including climate change, security challenges and the growing power of large technology corporations. His key message was that AI should not be seen as an autonomous force, but as a tool that can be used either for good or for harm, depending on who controls it and under what rules.
From a human rights perspective, Martin emphasised the importance of binding law. Ethical self-regulation by companies is not enough, especially when powerful technologies are deployed at scale. Human rights law offers an essential framework because it sets limits, protects fundamental freedoms and provides mechanisms for accountability. This is particularly important in areas such as surveillance, autonomous systems and high-risk AI applications.
The second part of the session focused more directly on the role of the European University Institute within ELOQUENCE. In the project, EUI leads the assessment and validation of AI outputs developed through different pilots. These assessments are carried out through an iterative process, involving multidisciplinary panels that examine whether the technologies being developed are aligned with human rights, ethics and European values.
Several concrete examples were discussed. One concerned privacy in smart home assistants. While such systems may be useful for supporting daily life, they also raise serious concerns when they listen to conversations within families or households. Privacy cannot be understood only as a relationship between one user and one device; it must also account for the different people, relationships and power dynamics within a home.
Another example focused on bias detection. Martin warned that bias should not be limited only to traditional protected characteristics such as gender, age, disability or ethnicity. AI systems may also create or reveal new grounds for discrimination, such as lifestyle patterns or personal traits that would not normally be considered legally relevant.
The webinar also explored emotion detection and emulated empathy. While detecting emotions may appear useful in some contexts, it also carries risks of misunderstanding, manipulation and harm. AI systems may misread cultural communication styles or use emotional signals in ways that disadvantage users. Similarly, AI that appears empathetic can mislead people, create false trust or even become dangerous in vulnerable situations.
The session offered a clear reminder: trustworthy AI is not only about performance. It is about rights, accountability, transparency and the careful recognition of human vulnerability.
Watch the full webinar here.
