Rethink AI Trust and Move From Compliance to Confidence

Every day, artificial intelligence becomes more woven into our decisions, in hospitals, ports, newsrooms, and beyond. Yet despite the promises of transparency and accountability, one question remains stubbornly open: do we actually trust AI?

The trust gap in 2025

The conversation around “trustworthy AI” has matured since the early ethics guidelines of 2019. We now have legislation, the EU AI Act, and a growing ecosystem of standards and certification schemes help with AI compliance. Still, a crucial gap persists between technical compliance and human confidence.

People don’t experience AI through checklists. They experience it through outcomes: a diagnosis they can’t understand, a headline they don’t believe, or a decision that feels opaque. In this lived reality, trust is not a box to tick, it’s a relationship to build.

From explainability to empathy

THEMIS Trust argues that assessing AI trustworthiness must start from users’ values and contexts, not only from developers’ intentions. A trustworthy AI system is one that people can reason with, not merely one that explains itself. This distinction matters. Explainability tells us how an AI made a decision. Trustworthiness tells us whether we should rely on it. That leap, from understanding to confidence, depends on whether the system aligns with human expectations of fairness, safety, and respect.

A risk-based approach to trust

Across THEMIS’s pilots in the port, healthcare, and media sectors, we see that trust can’t be standardised, but it can be systematised. By using a risk-based methodology, THEMIS helps organisations identify where trust might break down — and how to repair it. This means looking beyond bias detection or model accuracy to the full socio-technical landscape:

• Who is affected by the system’s decisions?

• What values are at stake?

• Which risks are acceptable, and which are not?

• How do we ensure that AI supports human judgement rather than replacing it?

  
  

Through structured assessments and co-design with workers and stakeholders, THEMIS builds trust that is context-aware, negotiated, and transparent.

Why AI compliance matters now

As generative AI enters workplaces and public services, the debate is shifting from “Can we trust AI?” to “How can we govern it responsibly in use?” THEMIS is part of that evolution. Instead of trying to make people adapt to AI, it helps AI adapt to people.

Trust in AI will not come from regulation alone, it will come from practice: from developers who design for dialogue, organisations that prioritise accountability, and users who feel empowered to question and understand.

In short, trustworthy AI is not a product. it’s a process.