Criticism of Model-Centric AI Regulation as Obsolete

Tech policy expert Seb Krier has sparked a debate by arguing that current AI regulatory frameworks are fundamentally obsolete because they rely on outdated technical assumptions. Krier asserts that because modern AI capabilities are increasingly driven by post-training, scaffolding, and inference scaling rather than just raw model weights, legal focus on the model provider is misplaced. He argues that mitigations such as safety filters and classifiers are typically applied at the software application layer, making model-centric laws ineffective. The critique suggests that regulators should pivot toward outcomes-based regulation and domain-specific oversight. Krier further contends that horizontal, entity-based approaches—similar to the European Union's strategy—fail to account for the complex chain of actors involved in modern AI deployment. He concludes that effective regulation requires deep vertical expertise rather than broad, all-encompassing interventions that ignore existing legal gaps and negligence standards.