Comment by Urban Institute

Organizations implementing responsible agentic AI must institute life-cycle governance: assigning clear human accountability and responsibility as well as mapping risks, measuring them, and managing them throughout design, development, deployment, and monitoring. Builders of responsible high-stakes agentic AI must implement clear ownership and decisionmaking pathways before any deployment. [...] If human oversight is called for at certain stages in the process, such oversight must be meaningful; “oversight” that merely approves the agent output without review is insufficient.