AI agents need ‘licenses’ to act, study says

Artificial intelligence systems that can act on their own should be registered and monitored like licensed professionals, according to a new proposal for AI governance. The idea, called “agentic profiles,” treats advanced AI not as a simple tool but as an independent actor—an “agent”—that makes decisions and takes actions without direct human control. Examples include AI that manages a supply chain, negotiates contracts, or operates a customer service line. The proposal argues that current rules focus on the people who build or sell AI. That is not enough. When an AI acts on its own, it should have a clear, verifiable identity that shows what it is allowed to do, who is responsible for it, and how it can be held accountable. Under this system, each AI agent would receive a profile—similar to a license or a passport—that lists its purpose, its limits, and its owner. This profile would be checked before the AI is allowed to operate in high-stakes areas like finance, healthcare, or public infrastructure. Supporters say this approach closes a critical gap. If an AI makes a harmful decision, regulators and the public need to know exactly which system did it, under what authority, and who must answer for the outcome. The proposal does not call for a ban on autonomous AI. Instead, it aims to make such systems transparent and traceable, so that governance keeps pace with technology that is already acting in the world.