Sign-Language AI Avatar Showcased at EU Parliament

Brussels conference examines how artificial intelligence can improve accessibility and healthcare while protecting dignity and fundamental rights

An artificial intelligence-powered avatar capable of communicating in Italian Sign Language and International Sign was presented at the European Parliament in Brussels during a conference examining accessibility, precision medicine and responsible innovation. Participants called for a European approach in which technological progress is measured not only by efficiency, but also by its ability to remove barriers, respect fundamental rights and improve people’s daily lives.

The international conference, titled Artificial Intelligence for Accessibility – From Innovation to European Standard, was held at the European Parliament on 14 July.

Representatives of EU institutions, scientific organisations, businesses and disability associations discussed how artificial intelligence could contribute to a more inclusive European digital environment. The central question was how practical innovations can be developed into reliable standards that public authorities and private service providers can use across Europe.

Accessibility placed at the centre of AI policy

The initiative was promoted by three Italian members of the European Parliament: Pietro Fiocchi, vice-chair of Parliament’s Committee on the Environment, Climate and Food Safety; Antonella Sberna, a vice-president of the European Parliament; and Chiara Gemma, a member of Parliament’s Disability Intergroup.

The opening session also included Alessandro Chiocchetti, secretary-general of the European Parliament; Alejandro Moledo del Río of the European Disability Forum; and MEP Francesco Torselli.

The choice of accessibility as the conference’s starting point reflected a wider change in European digital policy. The European Accessibility Act, whose requirements began applying in June 2025, introduced common accessibility obligations for a range of products and services, including ticketing systems, banking services, electronic communications, websites and mobile applications.

Accessibility is therefore becoming more than a voluntary addition to digital systems. It is increasingly being treated as a requirement that should be considered from the earliest stages of research, design and procurement.

An avatar communicating through sign language

One of the most closely watched demonstrations was an intelligent avatar developed by QuestIT, part of Vection Technologies. The digital assistant has been designed to understand and produce Italian Sign Language, commonly known as LIS, and to support communication through International Sign.

The system is intended to help deaf users obtain information and interact with digital services through a visual language rather than having to depend exclusively on written text, spoken announcements or assistance from hearing staff.

Unlike a prerecorded information video, a conversational avatar can respond to different questions and guide users through a digital service. Its potential applications include transport information, public administration, utilities, emergency services and other situations in which immediate communication may be necessary.

A version of the technology is already being piloted at the ticket offices of Roma Termini and Milano Centrale railway stations. The Trenitalia project was developed with the Ente Nazionale Sordi, Italy’s national association for deaf people.

According to information released by Trenitalia and the association, 30 sign-language users contributed approximately 8,000 recorded videos to the development and training of the system. That participation is significant because sign languages depend not only on hand movements, but also on facial expression, body position, rhythm and context.

Speakers at the Brussels conference also pointed to applications at Cagliari Airport and by the Italian water utility Acquevenete, where accessible digital tools can help users understand bills and obtain information about services. Further uses in public administrations and European institutions are being considered.

Technology should complement human support

AI-supported sign-language tools could make information available outside the hours in which interpreters or specialised staff are present. They may also help organisations provide routine information more consistently across busy transport centres and online platforms.

However, such systems must be developed with caution. Sign languages are complete natural languages with their own grammar and cultural context. Automated translation can produce errors, particularly when dealing with complex, legal, medical or emergency information.

Digital avatars should therefore not be treated as automatic substitutes for qualified human interpreters. Users must be able to identify when they are interacting with an automated system, obtain human assistance when necessary and challenge inaccurate or harmful information.

The European Disability Forum has recognised that artificial intelligence could significantly improve the lives of persons with disabilities, while warning that poorly designed or insufficiently tested systems can create new barriers. Disability organisations have consequently called for accessibility, safety and participation to be built into AI governance from the beginning.

This means involving deaf people and other intended users throughout development, testing and evaluation—not only after a product has been completed. It also requires clear rules on personal data, camera use, accuracy, cybersecurity and responsibility when a system fails.

AI, health and precision medicine

A second conference session examined the use of artificial intelligence in healthcare and precision medicine. Discussions focused on how data analysis and advanced computational tools could support prevention, diagnosis, medical research and treatments adapted to individual patients.

Participants included Antonio Novelli, director of the Medical Genetics Laboratory at Bambino Gesù Children’s Hospital and a professor at UniCamillus; Sergio Daga, a genetics researcher at Bambino Gesù; Salvatore Corrao, professor of internal medicine at the University of Palermo; and Stefano Crisci, who teaches artificial intelligence law at Sapienza University of Rome and serves on the board of Diplomatia.

European Commissioner for Health and Animal Welfare Olivér Várhelyi joined the accessibility discussion by videoconference. Other contributors included Elisa Zambito Marsala of Intesa Sanpaolo; Angelo Raffaele Cagnazzo of the Ente Nazionale Sordi; Gianmarco Biagi, president of the Italian Association for Business Culture and Management; QuestIT chief executive Ernesto Di Iorio; and Valter Mavrič, director-general of the European Parliament’s Directorate-General for Translation.

Elvira Tarsitano, a member of the national council of Italy’s federation of professional biologists, also participated.

The health discussion highlighted both opportunity and responsibility. AI can assist professionals in identifying patterns across large quantities of medical and genetic information, but its use may also affect highly sensitive personal data and important decisions concerning treatment and access to services.

Medical applications consequently require scientific validation, data protection, meaningful human oversight and clear accountability. A system may support a healthcare professional, but responsibility for decisions affecting a patient’s health and rights cannot simply be transferred to an algorithm.

From individual projects to European standards

The conference was coordinated by Carola Salvato, vice-president of Diplomatia and founder of KEA Connecta. It concluded with a round-table discussion on Europe’s approach to artificial intelligence and the need to combine innovation with responsibility, sustainability and respect for fundamental rights.

The debate comes as the EU moves from adopting artificial intelligence legislation to implementing it. The EU Artificial Intelligence Act establishes a risk-based regulatory system, while technical standards and enforcement mechanisms are still being developed.

For accessibility technologies, the move towards a European standard could offer important benefits. Common requirements could help public institutions compare products, demand independent testing and avoid buying systems that make impressive claims without providing reliable access in practice.

Standards should assess more than whether a digital avatar can technically produce signs. They should examine linguistic quality, comprehension, response accuracy, privacy, security, user satisfaction and the availability of alternatives when automated communication is insufficient.

The Brussels meeting showed that inclusive AI is no longer a theoretical possibility. Some applications are already operating in railway stations and other public-facing services. The next challenge is to ensure that expansion is guided by evidence and by the experiences of the people expected to use the technology.

Artificial intelligence can contribute to greater independence and participation when it removes a genuine barrier. Its success, however, will depend on whether Europe treats accessibility and human dignity as essential design requirements rather than optional features added after innovation has taken place.