The bipartisan discussion draft would establish new federal rules for AI agents that act on a user’s behalf, including requirements intended to promote user choice, security, and transparency.
Sen. Mark Warner (D-Va.) along with Sen. Todd Young (R-Ind.) released a discussion draft, The Artificial Intelligence Access, Gatekeeper Exchange, and Nondiscriminatory Transfer (AI AGENT) Act, that would create what sponsors describe as the first federal framework focused specifically on autonomous AI agents rather than AI models more broadly. The proposal comes as AI agents capable of completing multi-step tasks, accessing online accounts, and making decisions on behalf of users are moving rapidly from experimental tools to commercial products.
The draft proposes a federal framework to ensure AI agents act in the interests of the people they represent and promote competition among agent providers.
Among its major provisions, the proposal would:
- Require large online platforms to allow users to choose qualifying third-party AI agents rather than limiting them to platform-owned assistants.
- Direct the Federal Trade Commission (FTC) to establish security, privacy, identity verification, and interoperability standards for qualifying AI agents.
- Create a federal certification or registry process for AI agent providers that meet those standards.
- Require certified AI agents to operate with a “duty of loyalty,” acting in the interests of the user rather than favoring developers, advertisers, or affiliated businesses.
- Require transparency around financial relationships or conflicts of interest that could influence an agent’s recommendations or actions.
- Establish requirements for AI agents to securely access online services without exposing user credentials or compromising account security.
The senators are seeking public feedback on the discussion draft before introducing formal legislation.
