EU institutions seem poised to formally approve the world’s first horizontal AI law in the coming weeks. The AI Act’s first testbed will be on Friday 2 February, when ambassadors of EU Member States are set to vote on the agreed text. The European Parliament should start its own formal process of approval in mid-February, with a committee vote expected for 13 February followed by a plenary vote in April.
While we are awaiting the much-anticipated outcome of the vote(s), the tech industry is focused on what will come after the measure’s final passage. Now that the EU will have a legislative framework for AI, making it work in practice should be the top priority for EU policymakers and regulators. As industry prepares for the implementation work in the months and years ahead, here are some issues that should guide policymakers’ as they consider next steps:
-
Companies of all sizes operating in Europe will need legal certainty, guidance and predictable implementation across the EU Single Market. Regulatory complexity is an obstacle to European technological competitiveness and innovation and potentially a disincentive to investment. Risk of diverging interpretations or application of the EU AI Act in the Single Market should be minimized, especially as Member States are thinking about appointing different lead authorities for the enforcement of the Act, ranging from consumer protection to telecoms regulators. Structures like the AI Board and the newly formed AI Office must work to ensure EU-level coordination, further exchange between regulators, and provide much needed implementation guidance to companies and authorities.
-
The EU AI Act will not exist in a vacuum. It will coexist with several other EU policies applying to AI, such as the GDPR, cybersecurity regulations, the newly adopted Digital Services Act, as well as sectoral regulations. As ITI highlighted in our Vision 2030 policy guide for the next EU mandate, an ambitious assessment of the new body of EU laws applying to technology is needed, with the aim of identifying and streamlining potential conflicts and overlaps. When it comes to AI, analysis and guidance on the interactions with GDPR is especially needed.
-
International coordination around key AI governance concepts remains imperative. AI is developed through global value chains and a common understanding of key concepts across jurisdictions and like-minded countries will be crucial to ensure availability of new technologies, facilitate trade and support the ability of companies to compete globally. For example, diverging definitions of advanced foundation models (GPAI models with systemic risk in the EU AI Act; dual-use foundation models in the US AI Executive Order) or interpretations of emerging risks associated with them, could complicate companies’ AI governance efforts. As countries around the world consider their approaches to AI governance, continued and robust EU engagement in global forums and processes like the G7 Hiroshima AI process and the EU-U.S. Trade and Technology Council will be fundamental to support alignment and regulatory compatibility.
-
Global standards should be leveraged to demonstrate compliance with the EU AI Act. There are many standardization activities that are taking place outside of Europe, such as within ISO/IEC 42001, that can be relevant for the AI Act. A solely regional approach to AI standards will have negative consequences for innovation: it could create divergence and fragmentation of global regulatory environments and it will decrease the ability of European companies to scale and compete globally.
-
Implementation of the new rules should go hand in hand with support for innovation. AI is an opportunity for our societies and economies: it can bring huge advances in medicine, research, productivity as well as enable technologies that can contribute to the EU’s sustainability goals. Public policy should equally focus on how to reap the benefits of this transformative technology. The recent AI innovation package from European Commission focuses on the key role the EU can have in supporting AI innovation, as originally laid out by the goal of creating an ‘ecosystem of excellence’ in the 2020 AI White Paper. Together with supporting the implementation of the AI Act, the next EU mandate should continue this work and explore what the EU can do to boost the uptake of AI in Europe and the development and deployment of innovative AI use cases.