Operational responsibility in AI governance: a user-centric liability framework

Abstract Who bears responsibility when artificial intelligence systems cause harm? This question has become central to AI ethics and governance [55, 80]. Most existing approaches focus on developers, yet this faces serious practical and theoretical problems. Drawing on tort law, agency law, and philosophy of technology, this paper argues that AI should be understood as an instrument whose outputs remain the responsibility of human operators rather than developers. We call this 'user-centric governance.' Placing accountability with deployers promotes public trust by creating clear lines of responsibility, a concern that governance approaches have often overlooked [78]. It preserves democratic accountability by keeping human actors answerable for AI-mediated decisions, and it counters power imbalances by ensuring that those who use AI bear consequences for how they use it. Legal principles of instrumentality show that operational responsibility should follow use rather than creation. We propose a 'distributed yet centered' governance model that acknowledges developer obligations while treating deployment decisions as the center of primary accountability.