Building Trustworthy AI Together: Highlights from the THEMIS 5.0 Plenary in Valencia

Last week, partners of the THEMIS 5.0 project gathered in Valencia for the project’s 4th Plenary Meeting, hosted by Fundación Valenciaport. Over two days, researchers, technologists, and policy experts came together to review progress, align on next steps, and strengthen collaboration as the project moves into a crucial phase of integration and piloting.

Team THEMIS at Fundación Valenciaport

The meeting brought together consortium members working across sectors including ports, healthcare, and media, all united by a shared goal: helping organisations better understand and assess the trustworthiness of AI systems.

A Setting Focused on Collaboration

Held at the Fundación Valenciaport headquarters, the plenary provided an opportunity for partners to reconnect in person and engage in hands-on discussions. With the Mediterranean port as a backdrop, conversations ranged from technical architecture to policy impact, reflecting the interdisciplinary nature of the project.

Keeping Humans in the Loop

The meeting opened with a keynote on human oversight in the age of AI, exploring how organisations can maintain meaningful human control when deploying AI systems.

The talk highlighted an important challenge facing many sectors today: ensuring that humans are not merely formal overseers of automated decisions, but remain capable of understanding, questioning, and guiding AI-driven processes.

This theme resonated strongly with the broader mission of THEMIS 5.0, developing tools that support human decision-makers in assessing AI risks and trustworthiness.

From Framework to Platform

A central focus of the plenary was the integration of the THEMIS 5.0 platform, which combines several technical components designed to support trustworthiness evaluation. Partners presented progress on modules including:

• Tools that assess fairness, robustness, and safety of AI models

• Components that analyse user preferences and trust requirements

• Risk modelling and simulation tools for complex socio-technical systems

• Decision-support engines that help translate technical assessments into actionable insights

  
  

Together, these components form an ecosystem designed to help organisations evaluate AI systems not only technically, but also from ethical, social, and governance perspectives. The consortium also discussed the importance of ensuring the platform remains practical and usable for real-world contexts, a key priority as development progresses.

Testing in Real-World Scenarios

Another major topic was the preparation of pilot deployments, where the THEMIS tools will be tested in three different environments:

• Port operations, supporting trustworthiness evaluation of logistics and scheduling AI systems

• Media environments, helping assess AI used for detecting misinformation and harmful content

• Healthcare-related contexts, where trust, safety, and transparency are especially critical

  
  

These pilots allow partners to test the platform with real users, refine the tools, and gather insights about how organisations interact with AI assessment processes.

The upcoming months will focus on component integration, system testing, and iterative improvements, leading toward a validated platform.

Ethics, Impact, and Sustainability

Beyond technology, the meeting also explored legal, ethical, and societal considerations. Partners discussed how the project aligns with emerging regulatory frameworks such as the EU AI Act and how to translate research results into practical recommendations. The consortium is also preparing several outputs to maximise impact, including policy briefs, scientific publications, pilot case studies and collaborative dissemination activities.

These efforts aim to ensure that the THEMIS’s findings contribute not only to academic knowledge but also to policy discussions and industry adoption.

Looking Ahead for Trustworthy AI

With the project entering its final development phase, the Valencia meeting provided an important moment to align strategy, refine integration plans, and strengthen collaboration across partners.

Over the coming months, the consortium will focus on delivering a working platform, validating it through pilots, and translating project insights into guidance for organisations navigating the rapidly evolving AI landscape.

If the discussions in Valencia made one thing clear, it is that trustworthy AI cannot be built in isolation. It requires collaboration across disciplines, sectors, and perspectives — exactly the kind of cooperation that THEMIS 5.0 continues to foster.