Digital Ethics and Global Dialogue: Navigating the EU AI Act through a Multicultural Lens
Technology has never been neutral. But artificial intelligence; because it learns from us; risks magnifying our blind spots instead of our wisdom.
As we mark the World Day for Cultural Diversity for Dialogue and Development, the United Nations has reminded us that technology can either become a bridge between cultures or a tool for unintentional cultural erosion. The difference depends on one thing: who sits at the table when the rules are written.
At SKRC, we have taken that warning to heart. Our work bridging Nordic research excellence with international demand; from the Gulf to Southeast Asia; has taught us that AI ethics cannot be a western monologue. It must be a multicultural dialogue.
The Double‑Edged Sword
The 2026 UN theme for this day focuses explicitly on how digital transformation can either promote unity or deepen divides. AI is the most powerful example. When an algorithm is trained predominantly on data from one region, one language or one cultural framework, it does not simply perform poorly elsewhere – it actively imposes that original perspective as universal truth.This is not theoretical. Studies have shown that facial recognition systems trained on predominantly light‑skinned faces have error rates up to 34 times higher for darker‑skinned individuals. Large language models reflect the values, idioms and historical narratives of their training data – often English‑language, western sources. When these models are deployed globally, they do not translate neutrally. They export a worldview.
I do not say this to blame any single country or company. I say it because SKRC operates at the intersection of very different research cultures – Swedish, Nordic, Middle Eastern, Asian – and we see the mismatch every day. A Swedish AI ethics framework, built on GDPR and local social norms, may be excellent for Sweden. But it cannot be simply copy‑pasted to Abu Dhabi or Kuala Lumpur without asking: what is missing here? Whose perspective is not in the room?
What Our Operational Framework Demands
In our operational framework, we have explicitly integrated international R&D risks, including the ethical requirement for diverse data sets. Under the EU AI Act (2024/1689), which became applicable in 2026, our AI‑driven Matching Engine is classified as 'Limited Risk', meaning we must provide algorithmic transparency and allow human oversight.
But we went further. We commit to biannual conformity assessments. If our engine ever moves into 'High‑Risk' categories, we will implement mandatory Human‑in‑the‑Loop oversight, rigorous data bias logging, and full technical documentation as required by the AI Act's Chapter 2.
Why? Because a matching engine that recommends Swedish researchers to international funders could inadvertently favor certain universities, certain nationalities or certain research traditions. That would not only be unethical – it would be bad science. The best researcher for a given question might be a postdoc at a mid‑tier university with an unconventional background. Our algorithm must find that person, not default to the usual names.
This requires training data that is itself diverse. We cannot build a fair matching system using only data from established networks. We must actively include researchers from emerging economies, from minority backgrounds, from institutions that have historically been excluded from global collaboration. That is not charity. That is engineering.
But we went further. We commit to biannual conformity assessments. If our engine ever moves into 'High‑Risk' categories, we will implement mandatory Human‑in‑the‑Loop oversight, rigorous data bias logging, and full technical documentation as required by the AI Act's Chapter 2.
Why? Because a matching engine that recommends Swedish researchers to international funders could inadvertently favor certain universities, certain nationalities or certain research traditions. That would not only be unethical – it would be bad science. The best researcher for a given question might be a postdoc at a mid‑tier university with an unconventional background. Our algorithm must find that person, not default to the usual names.
This requires training data that is itself diverse. We cannot build a fair matching system using only data from established networks. We must actively include researchers from emerging economies, from minority backgrounds, from institutions that have historically been excluded from global collaboration. That is not charity. That is engineering.
The Professor's Privilege as a Cultural Mirror
Sweden operates under a unique legal framework called Lärarundantaget – the Professor's Privilege. Under the Act on the Right to Employee Inventions (SFS 1949:345), researchers own their inventions, not their universities. This is unusual in Europe, where most countries moved to university ownership decades ago.For an international funder from the Gulf or Asia, this is bewildering. They ask: "If I pay for research, who owns the result?" The answer requires explaining not just Swedish law, but Swedish culture – a deep respect for academic freedom, a distrust of corporate capture, a tradition of individual scholarly autonomy.
We designed our IP Bridge specifically to navigate this. It is a contractual framework that gives funders commercial certainty while protecting researchers' academic rights. But the deeper lesson is this: intellectual property is not just a legal concept. It is a cultural one. Different societies have different intuitions about who should own knowledge. Any ethical AI system that manages IP across borders must be trained to understand those differences, not flatten them.
A Continuing Dialogue
This is not a problem we will solve with one algorithm or one policy. It is an ongoing conversation – one that requires researchers, technologists and policymakers from every region to speak honestly about their assumptions and their blind spots.
At SKRC, we remain committed to hosting that conversation. Through our Swedish Dialogue Forum, our multilingual research index and our growing network of partners across the GCC, Europe and Asia, we will keep asking the hard questions: Whose data trains our models? Whose values shape our governance? And who is still missing from the room?
As the UN reminds us this week, cultural diversity is not a decoration on technology. It is a functional requirement. Algorithms that do not see the full range of human experience will produce results that are not only unfair – but wrong.
At SKRC, we remain committed to hosting that conversation. Through our Swedish Dialogue Forum, our multilingual research index and our growing network of partners across the GCC, Europe and Asia, we will keep asking the hard questions: Whose data trains our models? Whose values shape our governance? And who is still missing from the room?
As the UN reminds us this week, cultural diversity is not a decoration on technology. It is a functional requirement. Algorithms that do not see the full range of human experience will produce results that are not only unfair – but wrong.
#DoOneThingForDiversity.
We intend to build the right ones. And we invite you to join the dialogue.