New York City proposes requiring a human exit from AI customer-service loops

New York City wants businesses to offer an easy route to a human when automated customer service falls short. A rule proposed on October 20, 2026, would require disclosure of automated systems and meaningful help, backed by fines and restitution. It is a proposal, not an adopted requirement, with public comments preceding any final action.

From frustrating loops to consumer protection

The Department of Consumer and Worker Protection, known as DCWP, is targeting a specific pattern: businesses present customer service as available, then repeatedly fail to answer questions or supply irrelevant or wrong information. The agency proposes treating those failures as deceptive and unconscionable trade practices. Its approach focuses on the assistance a business actually delivers, rather than simply whether it has installed a chatbot.

Commissioner Samuel A.A. Levine framed the problem as more than an inconvenience. In the city’s announcement, he pointed to customers trying to refill critical prescriptions, rehabilitate student loans or recover large refunds. At a briefing covered by Gothamist, he said companies had used technology to ignore complaints and obstruct refunds, and warned that AI could worsen that relationship.

Disclosure would not be enough

The proposal pairs transparency with service obligations. Businesses would have to identify automated interactions, but labeling a chatbot would not satisfy the full rule. They would also need to provide timely and reasonable assistance and an easy way to reach a person if automated service falls short. The city describes three practical requirements:

• Tell customers when they are communicating with an automated customer-service system, including AI.
• Provide an easy way to speak with a human representative rather than trapping customers in automated exchanges.
• Deliver meaningful assistance when customer service is advertised, rather than repeatedly returning unresponsive, irrelevant or erroneous information.

The scope goes beyond text chatbots. Automated telephone voice-response systems are included, and Gothamist reports that the rule covers systems where technology handles all or substantially all communication. Levine said it would apply to companies doing business in New York City, including companies based elsewhere. A business’s headquarters would therefore not determine whether it falls within the proposed coverage.

The proposed rule also says businesses cannot evade responsibility for unlawful conduct involving their use of AI. It relies on the city’s Consumer Protection Law and would give DCWP citywide enforcement authority over businesses using automated customer-service systems. Financial consequences would include restitution to harmed consumers as well as civil penalties, with fines escalating for repeat violations.

$525First violation

The proposed civil penalty for a first violation.

$1,050Second violation

The proposed civil penalty for a second violation.

$3,500Third and subsequent violations

The proposed civil penalty for each third or subsequent violation.

November hearing comes before enforcement

DCWP published the proposal in the City Record on October 20, opening a 30-day public-comment period under the City Administrative Procedures Act. A public hearing is scheduled for November 19, according to Gothamist. The agency will review submissions before finalizing and adopting the rule, giving stakeholders a formal opportunity to respond to its requirements before they become binding.

The implementation date remains unsettled. Levine told Gothamist that timing would depend on the volume of comments the agency receives. He also said a transition period would allow companies to make the necessary changes after the rule takes effect. The hearing date is therefore not a deadline for businesses to have redesigned their customer-service systems.