How to Finalise the GPAI Code: A Test of Europe’s Commitment to AI Innovation
The finalisation of the General-Purpose AI (GPAI) Code of Practice represents a significant milestone in Europe’s journey towards establishing a robust framework for AI regulation. As the European Commission’s AI Office prepares to adopt the third draft of this code, stakeholders are increasingly concerned about the implications of regulatory overreach and the clarity of the text. This article examines the key challenges and recommendations surrounding the GPAI Code.
1. The Risk of Regulatory Overreach
The latest draft of the GPAI Code introduces several measures that extend beyond the legal framework established by the AI Act. These measures include:
- New definitions for open source AI models.
- Rights-reservation protocols allowing rightsholders to impose opt-outs from data mining.
- Ambiguous EU standardisation processes that could lead to regional fragmentation.
- Mandatory external assessments that had previously been rejected during negotiations.
This extensive list of new requirements raises concerns about the viability of the code as a voluntary compliance tool. If the measures do not align with the AI Act, signatories may find themselves at a disadvantage compared to non-signatories, ultimately undermining the purpose of the GPAI Code.
2. Clarity and Practicality Challenges
Another significant issue with the current draft is its persistent lack of clarity. Key deliverables from the AI Office, such as guidelines on GPAI rules and templates for public disclosure of training data, are still pending. The absence of these elements leaves the Code incomplete and difficult to interpret.
For instance, the introduction of concepts like ‘safe-originator model’ and ‘safely derived model’ requires alignment with forthcoming guidelines but currently lacks clear definitions. Furthermore, there are concerns about how trade secret protections will be maintained amidst new transparency requirements.
3. Ensuring a Strong Finish: Final Recommendations
The success of the GPAI Code hinges on three critical questions:
- Is it aligned with the AI Act?
- Is it clear and proportionate?
- Is it practical?
As the drafting process nears completion, these questions should guide the experts and EU Member States. It is crucial that the final Code supports companies in complying with the AI Act while also fostering an environment conducive to AI innovation and competitiveness within Europe.
Conclusion
The finalisation of the GPAI Code is not just a regulatory exercise but a test of Europe’s ambition to lead in AI innovation. A balanced, clear, and practical Code of Practice will empower companies to navigate compliance challenges while enhancing their capacity to innovate. As the EU moves forward, aligning its regulatory framework with the realities of the AI landscape will be essential for maintaining global competitiveness.
