AI Customer Service Rules: NYC Proposes a Human Off-Ramp
New York City wants to stop customer service bots from trapping people in endless loops. On October 20, 2026, the city’s consumer agency proposed a rule covering chatbots and other automated support tools. Violators would face restitution and fines that start at $525.
If you run a small team, you may have added a chatbot to save hours. That makes sense, because support is exhausting and hiring is hard. Still, this proposal is a signal that how your bot behaves is becoming a legal matter and not only a design choice.
What the City Actually Proposed
The Department of Consumer and Worker Protection published the proposal in the City Record. It would make it a deceptive practice to suggest help is available and then repeatedly fail to address a customer’s question. Unresponsive, irrelevant, or wrong replies would count too.
The agency groups its expectations into four duties. Businesses would have to tell customers when they are dealing with automation, provide an easy way to reach a person, and avoid blaming the technology for unlawful conduct. They would also have to deliver real help whenever they advertise it.
Fines start at $525, though the release states no maximum. It also says the rule builds on the city’s earlier click-to-cancel ban on subscription traps. In other words, this is part of a pattern of consumer-protection moves aimed at everyday business practices.
The rule covers automated systems broadly, including phone menus as well as chat windows. The release lists no size threshold or exemptions, so assume small businesses are included until the final text says otherwise.
Where Chatbots Fall Short
The city leaned on older research from Gartner to explain the problem. According to the release, bots resolve between 17 percent and 58 percent of issues depending on the task.
| Issue type | Resolved by chatbot |
|---|---|
| Billing disputes | 17% |
| Returns or cancellations | 58% |
Those gaps explain the frustration. A customer with a billing dispute is already stressed, and a bot that cannot help makes it worse. That frustration is what regulators are responding to.
The practical reading is that a bot is now part of your brand promise. If the promise is fast help, the bot has to deliver it or hand off gracefully. A rule like this simply writes that expectation down.
This Is a Trust Issue Before It Is a Legal One
Running a business is hard enough without a new rulebook. Even so, the underlying idea is plain: customers want to know a human will step in when a tool fails. If you have watched a support thread spiral, you know how fast goodwill drains.
Think about the last time a bot loop cost you time. You probably left that experience feeling less loyal to the company. Your customers feel the same way, and they rarely tell you before they leave.
Legal pressure on AI is growing in several places. We covered the fallout from an AI chatbot lawsuit recently. The direction of travel is consistent.
A Practical Audit for Your Support Flow
You do not need a lawyer to start. Walk through your own support experience as a frustrated customer and note where it breaks. Then fix the simple things first.
- Say clearly at the start of a chat that it is automated.
- Add a visible button or phrase that routes to a person.
- Log conversations that end without a resolution and review them weekly.
After the basics, measure. Count how many chats end with a human handoff and how long that takes. If handoffs are rare because customers cannot find the button, that is a design problem you can fix this week.
Consumer complaints can also escalate into bigger fights. Our story on mass arbitration claims shows how unresolved grievances can scale. A fast human response is cheap insurance.
One more note on tone. Rolling out an audit does not mean scrapping your chatbot. Many customers like quick answers to simple questions, and a well-built bot handles those well. The goal is a smooth path to a person for everything else.
What Happens Next
The proposal now enters a 30-day comment period that includes a public hearing. The city will then review submissions before it adopts a final rule. The release does not give a hearing date, so check the city’s rule page for the schedule.
Small firms can take part, too. The comment window lets owners explain what is workable, such as reasonable response times for a three-person team. Comments from real businesses can shape the final language.
If your customers live in New York, follow it closely. Other cities often copy first movers, so the habits you build now may save a scramble later. Federal attention is rising too, as our report on an FTC investigation shows. Be honest with your customers, give them a human option, and the rest gets easier.
