The AI AGENT Act is a discussion draft of federal legislation unveiled on June 29, 2026 by Sen. Mark Warner (D-VA), ranking member of the Senate Intelligence Committee, proposing a federal framework for consumer use of autonomous AI agents. The draft would create a registry of vetted "secure" AI agents administered by the Federal Trade Commission, establish security standards for agent products, and impose fiduciary-style duties on autonomous agents acting on consumers' behalf (Source: cyberscoop.com; cbsnews.com).
Status and legislative history
As of early July 2026 the measure remained a discussion draft circulated for comment rather than an introduced bill. Warner unveiled the draft on June 29, 2026; it became publicly reported around July 2, 2026 (Source: cyberscoop.com; insideaipolicy.com).
Key provisions
According to reporting on the draft, the measure would:
- Direct the FTC to maintain a federal registry of vetted "secure" AI agents, giving consumers a federally validated list of agent products;
- Establish security standards that agents must meet for listing; and
- Impose fiduciary-style duties on autonomous agents acting for consumers, addressing the interest-alignment problem that arises when an agent transacts on a user's behalf (Source: cyberscoop.com; cbsnews.com).
Warner said consumers need protection from AI agents as the products spread into everyday transactions (Source: cbsnews.com).
Context
The draft is one of several mid-2026 federal measures assigning new AI functions to the FTC, alongside the AI Labeling Act of 2026 (which Warner co-sponsors) and the Commission's enforcement role under the TAKE IT DOWN Act. It addresses the consumer-protection side of the agent-governance questions discussed at Agentic AI and Principal-Agent Problem Applied to AI, and follows the May 1, 2026 Five Eyes joint guidance on responsible agentic-AI adoption.
Commentary
A report from Stanford University's Institute for Human-Centered AI became public on August 26, 2026 describing the AI AGENT Act as an important but incomplete start toward a duty of loyalty for AI agents, and arguing for a fiduciary framework. The report itself could not be retrieved and the characterization is not verified beyond the trade-press headline summary (Source: insideaipolicy.com). See Stanford HAI, Principal-Agent Problem Applied to AI.
The American Fintech Council endorsed the draft in a letter dated 27 August 2026 and signed by chief policy officer Ian P. Moloney, released publicly on 1 September 2026. The letter names three provisions. Section 4(c) directs the National Institute of Standards and Technology to identify open protocols and, where none exist, develop model technical standards making online services accessible to custodial user agents. Section 4(h) preempts state law only to the extent it is inconsistent with the Act, and the council urged the committee to preserve that limited form. Section 4(g) directs the FTC to convene an interagency working group on fraud, misuse and error arising from custodial user agent activity. The council, which says it represents more than 150 member companies and partners, cited Colorado's Concerning Consumer Protections in Interactions with Artificial Intelligence Systems (Colorado AI Act (SB 24-205)) as an example of the state patchwork it opposes. Chief executive Phil Goldfeder said in releasing the letter that "it is critical that we establish rules to ensure they operate securely and fairly" (Source: fintechcouncil.org) (Source: insideaipolicy.com). The letter also supplies the draft's full title as the Artificial Intelligence Access, Gatekeeper Exchange, and Nondiscriminatory Transfer Act. The draft had not been formally introduced as of 1 September 2026.
Relationships
- related: Federal Trade Commission (FTC) — designated administrator of the proposed registry
- related: Agentic AI — the product category the draft would regulate
- related: Principal-Agent Problem Applied to AI — the fiduciary-duty provisions address this alignment problem
- related: AI Labeling Act of 2026 — companion Warner-backed FTC-enforced AI measure
- related: National Institute of Standards and Technology (NIST) — Section 4(c) would direct it to identify or develop agent-access standards
- related: Financial Services — AI Deployment — the sector whose trade body has endorsed the draft
- contradicts: Colorado AI Act (SB 24-205) — cited by the endorsing council as the state patchwork the draft's preemption clause addresses