Context: In April 2023, the European Commission (EC) made a formal proposal for a regulation on standard-essential patents (SEPs) (“EU SEP Regulation”). The European Parliament’s Legal Affairs Committee (JURI), which is the lead EP committee on this bill, is about to hold its vote, after which the measure will be put to a plenary vote.
What’s new: Unrelated to the process concerning the EU SEP Regulation, the full text of the EU AI Act became available today in the form of a four-column document juxtaposing the original EC proposal, the Parliament’s version, the Council’s stance and the compromise text that resulted from a “trilogue” (negotations between those three institutions). ai fray has commented on the AI policy implications of the proposal.
Direct impact: None as they are unrelated pieces of legislation.
Wider ramifications: At an abstract (meta) level, however, it’s worth talking briefly about striking parallels between the two proposals and the related processes.
The purpose of this article is not to discuss the AI-specific aspects of the AI Act in detail. In that regard, the above link to the related ai fray article should suffice.
Council as voice of reason
The legislative history of the AI Act shows that the European Parliament added to the Commission text a laundry list of partly very ideological and not always workable ideas. The Council had to ensure that the focus remained reasonably narrow on AI-specific questions, to prevent overregulation (though even the compromise text of the AI Act arguably goes too far) and to also ensure that the EU would not be overstepping its competencies in connection with certain use cases such as military applications.
The AI Act shows that the Parliament’s amendment-by-amendment voting system makes it particularly likely that different buzzwords make it into a bill. But regardless of procedural considerations, it’s clear that MEPs are more likely than the public servants in a Council working group to pursue political objectives even in a context where they don’t belong.
What the comparison of the institutions’ different positions also reflects is a tendency by the Parliament to overregulate and to focus too much on potential bad actors or abuse. In no small part, the Parliament’s version of the AI Act would have amounted to AI-related fearmongering.
Commissioner Breton is auditioning, therefore collecting fake achievements
Just like in the case of the SEP Regulation, internal market commissioner Thierry Breton’s primary interest is to position himself as someone who gets things done in the EU. He wants to be perceived as a shaker and mover (his staff even produced a short video for that purpose), and as someone who defends European interests against evil non-European companies.
He doesn’t seem to care about whether those proposals make sense. The number one priority is to get something done now, before the end of the term, so that he can add even more items to his list of claimed achievements. His key objective is to just position himself for higher office as the current EU term is already approaching its end.
Interestingly, in the case of the AI Act there are rumors of the government of his own country, France, trying to build a blocking minority in the Council against the trilogue result.
More eurocrats, more agencies
For the avoidance of doubt, the specific regulatory challenges that the AI revolution brings with it makes it a fairly reasonable proposal that the EU should have an AI-specialized team. In the AI Act, that one is called AI Office, but as ai fray explains in the article first linked to above, the European Parliament’s idea that there should be a separate agency was rejected by the Council and the compromise was to use the term AI Office as merely a synonym for the Commission’s AI-related efforts (i.e., for what will likely just be a Commission unit, or maybe even a directorate, but not a separate agency).
In the case of the envisioned Competence Center for SEPs, a similar compromise isn’t possible as the Competence Center would just be part of the EUIPO. But the word “competence” is a misnomer as they’d merely be outsourcing to external experts.
A compromise in the SEP context could be that the Competence Center would actually just focus on how to give advise to small and medium-sized enterprises (SMEs). That’s already the next topic:
SMEs as red herring
It looks as if the EU can’t do anything of industrial relevance anymore without at least the Parliament (if not the Commission in its original proposal) insisting on some special treatment of, and attention to, SMEs.
The AI Act contains various references to SMEs, as does the SEP Regulation, and one could mock this by saying the EU first creates regulations and then exempts SMEs from them in part or reduces fees. Neither the AI Act nor the SEP Regulation will make SMEs stronger. In an article published three days ago, Mattia Fogliacco, the president of (European) patent pool administrator Sisvel, explained how the SEP Regulation could even increase the risk of SMEs facing SEP assertions in the future.
Open source
Whatever the EU does with respect to the technology sector, there’s probably always someone who will argue that open-source software should also be treated differently. In the SEP Regulation, there could be a parliamentary proposal for some preference for open-source standards just because a member of the Green group from the German Pirate Party pushed for it. In the AI Act, it’s a far bigger problem because the regulation is intended to reduce risks, but open-source AI systems would lend themselves to abuse by bad actors, particularly criminals.
Failure to foster innovation and competitiveness
What ai fray criticizes about the AI Act is that it doesn’t do anything to foster innovation and make the EU economy more competitive. It’s just about regulation. In the case of the SEP Regulation, the Commission and some MEPs argue that it will be good for IoT SMEs and the automotive sector. As for the latter, it’s hard to see how an industry facing two major transformations in parallel (electric vehicles and ever more software challenges) will be “saved” by trying to reduce a per-car royalty of roughly $30. And with IoT SMEs, no one has so far been able to explain why they would suddenly be sued over SEPs if it hasn’t happened yet, at least not in a way that would have become known.
Inconsistent approaches to IP and standardization
Interestingly, the AI Act is much more positive about IP protection and collaboration with standardization bodies than the SEP Regulation.
Conclusion
Complex issues, particularly at the intersection of law and technology, are inherently more difficult for MEPs (than for the members of a Council working group) to figure out. MEPs don’t have experts on board. Their aides must deal with numerous policy topics all the time. But the Parliament plays an important role in the democratic process, provided that MEPs are realistic about the limits of their understanding.
A “trilogue” will always be about figuring out a face-saving path forward for the three participating institutions. In the case of the AI Act, many of the Parliament’s ideologically-driven ideas made it into the compromise text, but often in ways that won’t have much of a practical effect.
For the SEP Regulation it would be best to start all over again and explore completely different approaches. But if no political will can be built for that, the solution may just be to mitigate the impact. Many of the leaked questions by the Council reflect an awareness of a lot of issues with the proposed SEP Regulation, and the Commission wouldn’t be able to provide satisfactory answers in writing.
SEPs are a low-priority topic for the EU compared to AI. There’s a very strong desire to pass the AI Act into law before the end of the term. The SEP Regulation will apparently be carried over into the new term. At that point, Mr. Breton won’t be in charge. If the Council and the Parliament can’t work it out, the Commission might just withdraw the proposal. Unthinkable for the AI Act, but not for the SEP Regulation.