Talk to us 24/7: +1 (415) 650-2548 info@tendwiseai.com Free AI tools
Call us 24/7+1 (415) 650-2548 Book a Strategy Call

News · 5 min read · Published

New York City’s Proposed Chatbot Rule: What It Means for Your AI Customer Service

QUICK ANSWER

New York City has proposed a rule that would make any business serving New Yorkers tell customers when they're talking to a chatbot, give them a one-step way to reach a human, and keep three years of interaction records. Chatbots that trap people in "doom loops" would count as a deceptive practice, with fines starting at $525.

Black and brown headset resting next to a laptop on a desk

Key takeaways

  • The rule covers any business doing customer service with New York City consumers, not just companies based there, and includes phone voice systems as well as chat.
  • Bots must say they aren’t human at the start, and customers need a clearly labelled single-step route to a person, or another stated way to reach one.
  • Blaming the technology won’t work as a defence. Records of each interaction would have to be kept for at least three years.
  • It’s still a proposal. The city’s rules page now lists comments and the public hearing on 18 December 2026, so there’s time to check your bot before anything is final.

What New York City proposed

On 20 October 2026 the city’s Department of Consumer and Worker Protection (DCWP) announced a proposed rule aimed at what it calls AI customer service “doom loops”: bots that keep sending you back to the same menu, give wrong answers, or never let you reach anyone. Gothamist reported that the administration describes it as the first rule of its kind in the US.

It applies to any “automated customer service system”, meaning technology that handles all or substantially all of the conversation instead of a person. That covers website chatbots, in-app assistants and automated phone lines. And it applies to any company doing business with New York City consumers, wherever that company is based. A Dubai or London e-commerce brand shipping to Brooklyn is in scope.

The main requirements

Here’s what the proposed rule text asks for, in plain terms:

AreaWhat the proposal says
DisclosureTell the customer clearly, at the start and in the language being used, that they aren’t talking to a human.
Reaching a personOffer a clearly labelled single-step transfer (a button, a voice command, “Press 0”) with an estimated wait, or, if no one is available, give another route such as a phone line, live chat, email or form with hours or a response time.
Doom loopsPromising customer service and then repeatedly routing people to useless menus, or giving unresponsive, irrelevant or wrong answers, counts as deceptive and unconscionable.
No blaming the botA violation isn’t excused because a technology said or did the wrong thing.
RecordsKeep logs of each covered interaction, system details, copies of disclosures and annual aggregate data for at least three years, in machine-readable form.
Penalties$525 for a first violation, $1,050 for a second within two years, $3,500 for a third or later, plus restitution to harmed consumers.

A bot with no human option at all would be presumed to break the rule. That presumption can be rebutted, but you’d rather not be the one making the argument.

Why this matters beyond New York

DCWP Commissioner Samuel Levine told Gothamist he sees the rule as a model for other cities and states. That’s plausible. Gothamist, citing the Future of Privacy Forum, reports that legislators in 37 states have introduced over 120 chatbot bills, though most focus on disclosure and child safety rather than customer service.

There’s also existing law to think about. A joint brief from Protect Borrowers and Columbia Law School’s Center for Law and the Economy, written by two former regulators, argues that many chatbots may already break state rules on unfair and deceptive practices. It points to Air Canada, which a Canadian tribunal held responsible for its chatbot’s wrong answer about bereavement fares. That’s an advocacy brief, not a court ruling, but the Air Canada case is a useful reminder: your bot speaks for your business.

Our view: if you build for the New York standard now, you’re probably covered for most of what’s coming elsewhere. The same habits also help with the UK and EU push on transparency, which we cover in our UK and Europe work.

A practical checklist for your chatbot

  1. Check the opening message. Does it say plainly that the customer is chatting with an automated assistant? In every language you support?
  2. Add a visible “talk to a person” option that works at any point in the chat, not only after three failed answers. Show the expected wait or your human hours.
  3. Test for loops. Run your top 20 customer questions, including cancellations, refunds and complaints, and note any that end in a dead end.
  4. Turn on logging you can actually export, and decide who owns retention. Three years is longer than many chat tools keep by default.
  5. Review what the bot is allowed to promise. Refund amounts, delivery dates and policy exceptions are where wrong answers cost money.

We wrote a fuller guide on designing human handoff in AI support chatbots, and if you run support on WhatsApp, Meta’s own rules apply too, as covered in our WhatsApp chatbot policy article.

What happens next

The original notice set the hearing and comment deadline for 19 November 2026. The city’s rules page now lists both on 18 December 2026, so check it before you plan around a date. The notice doesn’t give an effective date, and Gothamist reports there’ll be a transition period once it’s final. Details could still change after public comment.

This article is general information, not legal advice. If you serve New York customers at scale, talk to a lawyer about how the final rule applies to you.

If you’d like a second pair of eyes on your bot, our AI chatbot development team builds disclosure, handoff and logging in from the start, and we work with businesses across the US. You can book a strategy call to go through yours.

Frequently asked questions

Does New York City’s chatbot rule apply to businesses outside New York?
Yes, as proposed. It covers any business providing customer service to New York City consumers through an automated system, wherever the business is based.
Do I have to let every customer talk to a human?
The proposal expects a single-step transfer to a person or, when no one is available, clear instructions for another way to reach one, such as a phone line, live chat or email with a stated response time.
What are the fines under the proposed rule?
$525 for a first violation, $1,050 for a second within two years and $3,500 for a third or later, plus possible restitution to consumers.
Is the rule in force yet?
No. It’s a proposed rule. The city’s rules page lists comments and a public hearing on 18 December 2026, and no effective date has been set.

Sources

Get a realistic plan for your business

Book a free AI strategy call and we will scope your workflow with you.