The much anticipated “Draghi-report” published earlier this month has by now made its way into most parliaments, cabinet offices and boardrooms. The almost 400 pages of analysis and recommendations provide a stark warning for the EU to clean up its economic and industrial act.
The conclusions from the main report (section A) have already been covered in op-eds, policy briefs and essays, but section B of the report, which contains sector-specific in-depth analyses, still contains valuable nuggets of information.
Many of the sectors mentioned in the report are direct research areas of The Hague Centre for Strategic Studies. In this Draghi Report Series, we decided to ask our experts for their views on specific sections of this high-profile report.
For the 7th installment of our Draghi Report Series, we asked strategic analysts Sofia Romansky and Jesse Kommandeur for their take on the report’s section on Artificial Intelligence.
– Which policy recommendations do you think are the strongest, and why?
Jesse Kommandeur: The Draghi report underscores the challenges facing the EU in the realms of computing and AI, emphasizing the region’s need to enhance its technological infrastructure and innovation capacity. The report identifies a critical gap in the EU’s ability to compete on a global scale, particularly in the development and deployment of advanced AI technologies.
I support the idea to “Harmonise national ‘AI Sandbox regimes’ across all Member States to enable experimentation and the development of innovative AI applications”. This policy recommendation seems to be designed to foster an innovative climate akin to that of the US: where developers and startups can experiment and innovate without the constraints that started to govern AI development over the last couple of years in the EU. The proposal’s strength lies in its potential to significantly boost AI integration and innovation by offering developers the freedom to build and test AI without immediate real-world repercussions. This flexibility is crucial for keeping pace with rapid technological advancements and regain a competitive edge compared to the US and China.
Furthermore, this recommendation suggests to “develop simplified rules, particularly for SMEs, and enforce harmonized implementation of the GDPR in the Member States, while removing regulatory overlaps with the AI Act“, which is particularly compelling for two reasons in the European context. First, SMEs often struggle with the adoption of AI due to resource constraints and the complexity of both technology and regulations. By simplifying rules and tailoring guidelines to fit the unique needs of SMEs, the EU can facilitate a broader adoption of AI technologies across a wider range of businesses. This focus is essential, especially given the EU’s ambitious target of achieving a 75% AI adoption rate by 2030, a significant leap from the current 11%, which predominantly includes larger corporations. Secondly, the emphasis on streamlined regulations addresses a critical barrier to AI innovation within the EU. While the establishment of the AI Act is a pioneering move by the EU from a privacy perspective, there is a real risk that overly stringent regulations could stifle creativity and technological advancement. As a result, the EU risks losing more talent, startups, and investment.
Sofia Romansky: Before diving directly into the section on digitalization and advanced technologies, there is one strength of the report that is worth pointing out. In providing a big picture view of the EU’s key sectors, the report already makes links with digitalization and advanced technologies in discussions of other sectors. This way, for example, certain AI applications as well as their opportunities and risks are quite well elaborated in a sub-section within the section on energy. Such an approach is important to ensure that sectoral policies are complementary and take advantage of the opportunities afforded by rapidly developing technologies.
Within section 3.2 on Computing and AI, it is essential that not only is AI discussed alongside computing, and specifically the requirements for cloud computing, but also that matters of computing come before those of AI. In national and international discussions around AI, regarding both societal and military domains, conversations often tend towards thinking about AI applications, phrased as “We would like to use AI for X to achieve outcome Y”. However, there is rarely reflection on the extent of infrastructural or institutional change that would almost certainly be required to support any array of AI uses. From the immense amounts of data and energy that are required to train sophisticated generative AI algorithms, to the reskilling that would be needed to train people to implement AI responsibly, it is not just a matter of ‘using AI’.
With this in mind, two recommendations are particularly strong. First, the suggestion to “develop homogenous and mandatory EU rules for sensitive areas of cloud services” is key for contributing to the EU strategic autonomy, safety, and ability to implement AI systems in alignment with EU values. While it is quite unlikely that all, or even a handful of, EU member states would agree to procure cloud services from a single provider, alignment in the processes of procurement can set minimum requirements to verify their security. This recommendation also comes with the sober acknowledgement that the EU is unlikely to be able to compete with US cloud providers. However, in terms of integrating AI technologies as a whole, the EU could still possibly catch up if it were to “Develop and fund a strategy to rapidly enhance the EU’s computing infrastructure and AI capabilities, connect private and public computing nodes, and reinvest returns of this public ‘computing capital’ in new capacity.” Especially valuable is the investment in private and public linkages. Given the dispersed nature of AI model and algorithm development, the pooling of resources as well as best practices is part of an optimal solution.
– Is there anything missing in the policy recommendations? What would you add?
Jesse Kommandeur: As mentioned earlier, while harmonizing regulations and creating supportive frameworks for AI development are relevant, the priority should not solely be on additional high-level policy, legislation and frameworks. Instead, the focus should increasingly shift towards implementing practical incentives that enhance the EU’s attractiveness as a hub for AI talent and innovation, such as more aggressive fiscal incentives for investments in startups and tech firms, especially in cutting-edge sectors like AI and quantum computing. Or investing in universities to bolster research and development capabilities could cultivate a more robust talent pipeline for students and graduates.
Sofia Romansky: While not directly related to the policy recommendations, it is necessary to be wary of not feeding into the hype surrounding AI as something that will undoubtedly contribute to economic prosperity. The potential fallout of job loss and impact on education and knowledge accumulation remains to be seen. Simultaneously, given the costs associated with significant infrastructural and institutional changes that will need to take place before AI is comprehensively integrated in EU systems and sectors, it may still take some time before it becomes possible to reap the benefits. Finally, when the report suggests to “Identify priority AI vertical applications for the EU, encouraging EU companies to participate in their development and deployment in key industrial sectors”, it is important that this is done in a way that avoids encouraging ‘innovation for innovation’s sake’ or the Silicon Valley mindset of ‘move fast, break things’. As outlined by the report itself, the EU does not operate in the same innovation space as the US. But, by fostering its own EU-wide innovation mindset, its companies can and should be purposeful in their innovation and aligned with EU values.
– Which figure or data point in the report did you find most insightful, and why?
Jesse Kommandeur: The differences in investment between the EU and US are stark. In 2023, the EU attracted a mere USD 8 billion in venture capital for AI, dwarfed by the USD, with USD 68 billion that flowed into the US and USD 15 billion into China. The substantial investment gap between the EU and its global competitors not only impacts the pace and scale of AI development today, but also has implications for the future. As this disparity widens, the long-term consequences may include increased technological dependence and innovation stagnation, which could severely restrict the EU’s technological sovereignty.
– How do you view the feasibility of these plans in an European context?
Jesse Kommandeur: The feasibility of these plans in a European context really depends on the ability and willingness of EU member states to effectively coordinate and harmonize AI development and its computational infrastructure. In addition, achieving integration across diverse national landscapes has its own challenges, for example due to varying HPC infrastructural properties or cloud-related privacy laws.
Moreover, even if the EU succeeds in implementing these changes, it remains uncertain whether the policy recommendations will sufficiently enhance the EU’s competitive edge. This skepticism is primarily due to the significant lead already established by the US and China, who not only dominate in technological advancements but also benefit from robust ecosystems that support continuous innovation and substantial investments in AI. Consequently, the EU must not only adopt and implement these policies effectively but also ensure they are complemented by strong support mechanisms, strategic funding, and incentives that can drive substantial advancements within its AI sectors.
Stay tuned for the next issue of the Draghi Report Series by HCSS, where we invite experts to discuss other sectorial policies in Mario Draghi’s report “The future of European competitiveness”.
Read the previous installments here:
- Draghi Report Series | Patrick Bolder on Space
- Draghi Report Series | Benedetta Girardi on Semiconductors
- Draghi Report Series | Han ten Broeke and Ron Stoop on Automotive
- Draghi Report Series | Irina Patrahau and Michel Rademaker on Critical Raw Materials
- Draghi Report Series | Tim Sweijs and Frank Bekkers on Defence
- Draghi Report Series | Lucia van Geuns and Jilles van den Beukel on Energy
Related Content
