Joel Kaplan says Meta won't sign EU's AI code of practice, calling its guidelines overreach that will throttle development and deployment of AI models in Europe
Meta Platforms Inc. said it won't sign the code of practice for Europe's new set of laws governing artificial intelligence …
Context & Ripple Effects
Meta's stance extends a prior pattern of friction over European AI rollout: it previously held back Meta AI in Europe after an Irish privacy regulator sought a pause on training with public user content the earlier delay to its European Meta AI launch.
The code was voluntary and intended to help firms align with the AI Act, while the US had already pushed back on its development US pressure over the voluntary code. Meta's refusal therefore tests whether voluntary implementation can build broad industry consensus.
First-order effects
- Meta will not use the code as its chosen voluntary compliance framework, leaving European regulators without its public endorsement of the guidelines.
- The move makes Meta's disagreement with Europe’s AI governance approach explicit, rather than limiting the dispute to the operational constraints around a single product launch.
Second-order effects
- Other AI providers can more visibly choose between alignment with the code and Meta's confrontational posture; Google’s subsequent decision to sign the code sharpens that contrast.
- A split in participation reduces the code’s value as a common implementation signal for companies building or deploying models across Europe, even though it remains voluntary.
Third-order effects
- If major model providers increasingly pursue different compliance paths, European AI access and deployment could become more fragmented by jurisdiction rather than governed through a shared industry baseline.
- The episode points to a durable contest over whether voluntary codes can bridge the gap between formal AI rules and rapid model deployment; that outcome remains dependent on enforcement and participation by other providers.
The trend: This is one instance of model-access balkanization, as major AI companies increasingly treat regional governance choices as product and deployment constraints.