Skip links

EUs Digital, AI and Data Strategy lacks ambition on ethics and Trustworthy AI

The strategy is ambitious on advancing the European AI uptake and taking back control over a European data space. But it lacks ambition when creating the conditions for a thriving ethical and trustworthy AI and data space.

On 19th February 2020 the European Commission published a comprehensive strategy on the digital, AI and data future of Europe. The strategy followed after a two-year period of scrutiny into the role of Europe in the global AI space with among others the creation of a set of European ethical guidelines for Trustworthy AI to inspire the policy process.

The new strategy consists of:

The key difference between the three is that while the AI white paper is open for public consultation and therefore concrete actions outlined in this are not set in stone,  the two communications are actual strategies. Nevertheless, in the following I look at the white paper and the digital and data strategy in combination as an altogether expression of the current direction of the European Commission’s European digital, AI and data strategy. It is important to remember here that the focus on creating a digital agenda for Europe is not a new policy topic in the EU (including a couple of communications published last year on European data sharing etc.). It has over the years been a negotiated topic invested with many stakeholder interests inside and outside the EU (See more below). Nevertheless, this is the first time that we see a more comprehensive data and AI strategy as such. In addition, the AI and data agenda has received great attention worldwide framed on the global scene as a particularly “European” response from the EU that increaingly is seen as a “regulatory super power” in the global digital space (See my analysis on this in this working paper which is still in review)

The AI regulation?

Especially the AI white paper was much anticipated after the new president of the European Commission Ursula von Leyen in late 2019 pledged that she would regulate AI within the first 100 days of her presidency. Many were therefore waiting for an AI regulation as such. But the white paper only considers mapping of existing law, possible adjustment and clarification of law and only new provisions in areas of AI adoption and development considered high risk are envisioned. Nicolas Kayser-Bril from AlgorithmWatch’s analysis of the white paper in particular in regards to the legal aspects and actions on high risk AI is here an essential read.

The policy and investment strategy

Now, one thing is the law, another is the overall policy and investment aspirations. The strategy has generally an important emphasis on the human centric approach and the concept of trustworthy AI and in the white paper even spells out the key requirements of a trustworthy AI system that we in the AI High Level Group had identified and included in the ethics guidelines published last year (Human agency and oversight,  Technical robustness and safety, Privacy and data governance, Transparency, Diversity, non-discrimination and fairness, Societal and environmental wellbeing, and Accountability).

This is a very positive and essential positioning of the digital, AI and data strategy as a whole. However, the very AI ethics and trustworthiness of a socio-technological development needs action, and here we should not only consider law, compliance and risk mitigation frameworks. Steering a development as such demands an additional combined set of governance tools, from investment in and policies around cross cutting research and science, standardization, infrastructural development, public procurement to innovation and business development. And I do not see this as part of the agenda.

In the current strategy, what has also been dubbed the values-based European “third way” on the global AI agenda is primarily addressed in a risk based compliance and requirements framework and the danger is that it will also consolidate in Europe as such. In this way, we are running the same risk as we did with data privacy in the past, where the management of privacy rights was placed on the shoulders of the users, fixing it as afterthoughts in processes of developing and adopting AI.

Ethics needs to come before development, deployment and adoption. But it does not come on its own. We do not have thriving ethical AI innovation yet, that is something we need to develop and that should be invested in. Computer science and engineering students do not consider it, developers are not aware of it. We are still at an early experimental stage of the development in which the technologies and methods that will enable this are scarce and still widely debated. We still need to develop concrete things like human in the loop technologies and organizational cultures, state of the art data minimization and anonymization methods, ways of assessing social and ethical implications etc. etc.

In the AI white paper the only concrete plans for creating the conditions for trustworthy data and AI innovation and adoption beyond a risk-based legal compliance framework, are:

  • Transforming the ethics guidelines’ assessment list for AI practitioners into a training kit
  • Promoting responsible data management practices


  • Promoting ethical principles and trustworthy AI in international fora

The data strategy is in regards to general AI uptake on the other hand very ambitious, all-encompassing, and includes all the governance tools needed to shape the European data and AI evolution. It describes the development of data driven eco systems, data pooling and sharing infrastructures, data interoperability, harnessing of data market power, strengthening Europe’s capabilities and infrastructures for hosting, processing and using data, interoperability and not the least the development of eu based cloud providers, ensuring AI uptake in public procurement, and creating common European agriculture public administration, industrial (manufacturing), Green Deal, mobility, space,  financial, health, and skills data spaces. However, also here the conditions for creating a values-based adoption and development of this European data eco system, beyond some good ideas regarding personal data control management systems, some skills development of engineers and what is referred to as the promotion of “data altruism” among users, only has a few dedicated concrete actions framed as requirements such as:

  • enhancing the portability right for individuals under Article 20 of the GDPR giving them more control over who can access and use machine-generated data
  • rules for providers of personal data apps or novel data intermediaries such as providers of personal data spaces.

These actions lack the ambition of the policy and investment plan enabling Europe’s general AI uptake. If we go back to the AI High Level Group’s policy and investment recommendations from last year, we actually made quite a few to ensure the realisation of trustworthy AI in the European general AI uptake, e.g. making strategic use of public procurement to ensure trustworthy AI (not just AI), develop legally compliant and ethical data management and sharing initiatives in Europe, funding and facilitating the development of tools to assist in detecting bias and undue prejudice in governmental decision-making, develop a cross-cutting network focused on Trustworthy AI across European universities and research institutions, foster the availability of redress-by-design mechanisms etc. etc.

All in all, while previous communications from 2018-19 put forward the values-based approach of Europe as not only a requirement, but importantly also an opportunity and competitive advantage, the new strategy  does not describe trustworthy AI per se as an opportunity and worth investing. In fact, it is presented as ancillary to the European AI uptake. But if Europe truly wants its adoption of AI to be different and competitive in its own unique way with the human centric approach being more than just a mantra, then we also need to make an honest investment in creating this type of market.

The politics and interests

This new strategy has one over-arching very important political goal, which is its core focus on creating a comprehensive data strategy for Europe. Our economic, political, social and cultural infrastructures are built on data. AI systems feeds on and evolves on data, and so the data of it is a resource invested with a lot of interests and power dynamics. Outside and inside Europe. The new strategy recognizes this in more than one sentence and statement. In Europe, we do not have a tradition within business and innovation for moving fast and breaking things along the way. The goal is not just growth, but sustainable growth, which also means that innovation has a certain pace. The rapid development we have seen over the years with very aggressive data harvesting by big technology companies from outside Europe and the disruptive democratic effects of this does call for a more protective stance. This is also why the European strategy calls for what is referred to as “European technological sovereignty” emphasizing that this starts from “ensuring the integrity and resilience of our data infrastructure, networks and communications.”

History and mindsets

With the force and public attention Europe’s new strategy has been presented with, it might be overlooked that it did not evolve in a policy vacuum. It is part of an ongoing European digital agenda narrative that builds on different historical political aspirations of the European Union in general. In the midst of this we need to remember the roots of these narratives and shape the direction that it will take in European societies going forward. For example, back in 2010 a Digital Agenda for Europe was presented with the description of a “Digital Single Market” stating: “It is time for a new single market to deliver the benefits of the digital era”. Within this agenda, the Digital Single Market aspiration was voiced in a concern over a persistent fragmentation that was said to be restraining Europe’s competitiveness in a digital economy overshadowed by companies, such as “Google, eBay, Amazon, and Facebook” that “originate outside the EU”. Following this, a number of policy communications and initiatives laid the foundation for a European competitive big data digital space with much of the overly positive language on the unquestionable opportunities and promises of big data that we also see now in 2020 in the new  Digital, AI and Data strategy. The general political aspiration to compete on the same terms as the giants from the outside has however changed over the years with an increasing focus on the societal and ethical implications of digital big data technologies and businesses. This happened alongside several controversial publicly debated cases revealing the more negative democratic effects and power dynamics of the new global big data infrastructures (from Snowden revelations to Cambridge Analytica). In response to this, notably a reform of Europe’s General Data Protection legal framework was implemented and a much stronger aspiration to create a European data infrastructure and digital single market as a differentiator on a global competitive digital market took form. This is also where a European values-based “third way” emanating from a European Fundamental rights framework really unfolded as an actual policy focus. In 2019, Ursula von Leyen summarised years of negotiation between two somewhat competing big data aspirations reconciling them in one aspiration: “In order to release that potential we have to find our European way, balancing the flow and wide use of data while preserving high privacy, security, safety and ethical standards. “

See also:

Data Ethics Forum: Join and discuss more at the Data Ethics Forum in Rome 29-30 May 2020 “Mapping Artificial Intelligence: The European AI Agenda in a Global AI Ecosystem”