EU lawmakers are on a tricky knife edge hammering out the AI Act's final shape, as France and Germany push for a regulatory carve-out for foundation models

TechCrunch 2023-11-15 Natasha Lomas

Context & Ripple Effects

The dispute sits between Parliament's earlier push for copyright-use disclosure rules for chatbot makers and the eventual inclusion of additional duties for some proprietary models deemed systemic risks. It shows that the Act's foundation-model regime was a central unresolved political question, not a late drafting detail.

The subsequent political deal pairing the Act with pro-innovation measures and an AI Office suggests the eventual framework would need to balance enforceable obligations with national concerns about European AI competitiveness.

First-order effects

France and Germany's intervention puts the scope of foundation-model obligations back into negotiation, creating immediate uncertainty for model developers over whether a single EU-wide rulebook will apply to them.
Lawmakers face a sharper choice between preserving broad model-level safeguards and accommodating an exemption sought by two major member states, potentially delaying agreement on the final text.

Second-order effects

A carve-out would pressure other member states and Parliament to seek compensating safeguards or narrower definitions, making compliance requirements more dependent on model classification and jurisdictional politics.
Developers and downstream companies would have less clarity on documentation, transparency and risk-management expectations until the treatment of foundation models is settled.

Third-order effects

If national competitiveness concerns repeatedly reshape horizontal AI rules, EU AI governance may evolve toward differentiated obligations for frontier-model providers rather than uniform obligations across the bloc.
The episode is an early test of whether the EU can institutionalize frontier-AI oversight while retaining political support from states seeking domestic model capacity; the later systemic-risk framework indicates compromise rather than a clean exemption may be the durable path.

The trend: Frontier-AI regulation is becoming a contest between common-market safeguards and member-state efforts to protect strategic domestic model development.

Discussion

@zach_cer Zach Meyers on x
I am never quick to buy into the argument that EU regulations are protectionist. But FR/DE doing themselves no favours by suddenly trying to weaken the AI Act to protect their own AI start-ups.
@vkhosla Vinod Khosla on x
If the EU does what @tegmark wants they will fall further behind. Way to go. Ignore top 5 AI risks to take care of nonsensical hard takeoff AI and deny humanity huge benefits.
@ai_ctrl @ai_ctrl on x
🚨 @MistralAI CEO Arthur Mensch does not deny “pressing lawmakers” but rejects the suggestion @MistralAI is “blocking anything” in the EU AI Act 🚨 Sure, they are not “blocking anything” - just gutting the whole Act by trying to exempt the most relevant & powerful AI systems 🙃 [ima…
@dr_park_phd Dr. Peter S. Park on x
Blatant falsehood from Facebook's Dr. LeCun. The French and German governments were lobbied by Mistral, Aleph Alpha, and OpenAI+US Big Tech to block the foundation-model requirements in their entirety, not just to exempt the so-called “open source” models. https://x.com/... [imag…
@josephjacks_ @josephjacks_ on x
“Lawmakers must stand firm and protect thousands of European companies & consumers from the lobbying and regulatory capture of @MistralAI and @Aleph__Alpha” *Actually* open source AI companies = the opposite of regulatory capture. Max has officially gone insane.
@stevesi Steven Sinofsky on x
EU's AI Act negotiations hit the brakes over foundation models - https://euractiv.com/ // imagine a world where the US rush to enable regulatory capture results in more regulation than EU. Be careful of what you ask for. https://www.euractiv.com/...
@bertuzluca Luca Bertuzzi on x
There is nothing better than some Friday drama to close such a hectic week. A technical meeting on the #AI Act has broken down today, but make no mistake: the issue is deeply political and, if no solution is found soon, the whole law is at risk. A 🧵1/8 https://www.euractiv.com/..…
@tegmark Max Tegmark on x
This last-second attempt by big tech to exempt the future of AI (LLMs) would make the EU AI Act the laughing-stock of the world, not worth the paper it's printed on. After years of hard work, the EU has the opportunity to lead a world waking up to the need to regulate these incr…
@krishnanrohit Rohit on x
@tegmark You can't lead the world in technology through premature regulations. This didn't even work for technology that's easily copyable, like financial services, much less for frontier technology like AI and LLMs.
@ylecun Yann LeCun on x
This is *not* “Big Tech” versus The People or whatever. This is open source AI versus closed and proprietary AI. On the one hand, you have Mistral, Aleph, HuggingFace, Meta, IBM, and the entire startup ecosystem arguing for open source AI foundation models. On the other hand, …

Chronicles