Comment on FTC Negative Option Rule ANPRM

ROSCA and the Commission's legacy Negative Option Rule rest on a foundational assumption: a human consumer reads disclosures, considers terms, and affirmatively acts to accept a recurring obligation.

"Express informed consent" under ROSCA requires that the consumer receive, and affirmatively agree to, material terms before being charged. The concept presupposes that a human being is reading, understanding, and deliberately accepting those terms.

"Clear and conspicuous" disclosure is defined in visual and auditory terms — prominence, proximity, placement, font size, contrast. These are design constraints for human perception. They are meaningless to an AI agent parsing a website's DOM structure or interacting via API.

"Simple mechanism for cancellation" assumes the consumer who enrolled is the same entity that must cancel. When enrollment is performed by an AI agent but cancellation requires a human to navigate a phone tree or multi-step web form, the symmetry principle is violated by design.

The Commission need not speculate about whether AI agents will conduct commerce on consumers' behalf. Multiple commercial platforms are already in market:

OpenAI ChatGPT Instant Checkout (2026): Enables ChatGPT to complete purchases within the conversational interface, including subscription sign-ups. The user delegates purchasing authority via general instruction.

Google Universal Commerce Protocol (2026): A technical framework for AI agents to transact with merchants via structured APIs, bypassing traditional web-based checkout entirely.

IAB Agentic Commerce Roadmap (early 2026): Explicitly envisions AI agents navigating advertisements, evaluating offers, and executing purchases — including promotional trials, subscription offers, and bundled add-ons.

Each of these platforms creates scenarios where traditional ROSCA safeguards fail:

Scenario 1 — Agent subscribes: Consumer set a general objective ("find a coffee subscription under $50"), not specific consent to the terms the agent accepted.

Scenario 2 — Visual terms: Agent interacts via API or DOM parsing. Visual prominence is irrelevant to a machine.

Scenario 3 — Free trial converts: Agent accepted trial terms; consumer may not know a negative option feature was included.

Scenario 4 — Consumer wants to cancel: Agent enrolled automatically; cancellation requires a human to navigate a multi-step process.

Scenario 5 — Dynamic pricing: Consumer never sees the actual terms the agent accepted.

Scenario 6 — Agent malfunctions: Unclear whether liability falls on consumer, agent developer, or merchant.

We propose a safe harbor framework for merchants who build AI-readable disclosures and agent-compatible cancellation mechanisms. Merchants who implement machine-readable disclosure formats (structured data, APIs) and symmetrical cancellation pathways (if an agent can enroll, an agent can cancel) would receive presumptive ROSCA compliance for AI-mediated transactions.

This approach creates market incentives for merchants to modernize their disclosure and cancellation infrastructure while providing the Commission with an enforceable framework for the emerging agentic commerce ecosystem.

The Eighth Circuit's vacatur of the 2024 Click-to-Cancel Rule in Chamber of Progress v. FTC — for failure to conduct the preliminary regulatory analysis required by Section 22 of the FTC Act — presents the Commission with a rare opportunity to rebuild the regulatory framework from the ground up. We urge the Commission to include AI agent purchasing in its Section 22 analysis, ensuring that the next iteration of the Negative Option Rule addresses the technology landscape as it actually exists.