Governing Actions, Not Agents: Institutional Attestation as a Governance Model for Autonomous AI Systems
Autonomous AI agents may begin to perform consequential, irreversible actions such as clinical prescribing and production software deployment. This paper observes that human institutions have governed powerful autonomous actors not by monitoring their reasoning but by requiring i...
ArXiv CS.AI
Source
Generated insight by OnAirToday AI
This paper proposes a novel governance model for autonomous AI by shifting focus from agent reasoning to external attestation of actions. This approach leverages existing institutional trust mechanisms to ensure accountability for high-risk AI operations, offering a practical path to safer AI deployment.
Get the day's AI news in one short email. Free.
Comments
Keep comments relevant and respectful. No links, advertising, self-promotion, or spam.