On June 14th, the European Parliament approved their negotiating position on the EU AI Act, the world’s first comprehensive law on AI, with an overwhelming majority. The bill now moves into trilogue negotiations—interinstitutional negotiations between the European Parliament, European Commission, and Council of the EU—and legislators are aiming to close the file by the end of the year.

The Future Society has been instrumental in shaping the future of AI regulation in the European Union. Recognizing the EU’s leading role in tech governance, TFS anticipated its pioneering actions in AI regulation. In 2018, we organized the Global Civic Debate on Governing the Rise of Artificial Intelligence in the European Parliament, and have provided consistent recommendations to key policymakers for the past 5 years.

Since the European Commission unveiled the draft Artificial Intelligence Act in 2021, we’ve been committed to ensuring that the Act and the overarching regulatory framework incorporate ample caution with respect to increasingly powerful AI systems. Our research has delved into the governance of general-purpose AI systems (GPAIS), informing EU regulators about their potential risks and advocating for governance mechanisms that promote trust, designing a phased roll-out scheme for regulatory sandboxes, and laying out future-proof policy measures, such as a navigator programme for compliance by design, involving a Code of Conduct for GPAIS research and development. Our work extends beyond legislative measures, as we actively strive to foster dialogues aimed at mitigating competitive race dynamics in the AI industry while promoting global coordination.

The draft adopted yesterday contains a number of measures for which TFS has been advocating. Collectively, these operationalize safety, fairness, accountability, and transparency in the development and deployment of AI systems. These include:

• A special governance regime for foundation models (Article 28b).
• The establishment of an AI Office with legal personhood, empowered to update guidelines on training thresholds for foundation models (Article 56).
• National, regional, and EU-wide regulatory sandbox systems (Article 53).
• The establishment of a direct complaint procedure (Article 68c).
• Whistleblower protections (Recital 84c).
• Benchmarking capabilities (Article 58a).

Having received overwhelming support from the Parliament, we are optimistic that the EU will adopt these measures. We are grateful to EU policymakers’ tireless work to build such a governance regime. We look forward to continuing our advisory and collaborations on these topics and to fulfilling our mission to align AI through better governance.

Read more about our work on European AI Governance.