---
title: "“Connect me to a human.” New York wants to make it mandatory."
description: "New York City wants to stop businesses from trapping customers in chatbot loops. The rule would only apply there, but the questions behind it matter to anyone running a chatbot or voicebot in customer service."
url: https://www.lautmaler.de/blog/chatbot-human-handover
lang: "en"
---
[Blog](https://www.lautmaler.de/en/blog) Law & Privacy

# “Connect me to a human.” New York wants to make it mandatory.

New York City wants to stop businesses from trapping customers in chatbot loops.

6 min read 2026-10-26

![Titelbild zum Artikel „Bitte mit einem Menschen verbinden. New York macht’s zur Pflicht.“](https://www.lautmaler.de/uploads/3o28mtUNE_Dgy1Wv-pL06-Blogbeitrag_Bitte_mit_einem_Menschen_verbinden.png)

## In short

On October 20, 2026, New York City’s Department of Consumer and Worker Protection (DCWP) proposed a rule for automated customer service. Businesses would have to disclose that a bot is answering, offer a one-step route to a human and stop sending customers in circles with irrelevant or wrong answers. Violations would cost between $525 and $3,500, and the comment period runs until November 19. In the EU, the disclosure duty under Article 50 of the AI Act has applied since August 2026, but it does not include a right to a human.

## What does New York’s proposed chatbot rule say?

The DCWP published the draft on October 20, 2026 in the City Record, the city’s official journal. If adopted, New York would be the first US city with the power to enforce a rule like this, according to the agency. It covers systems where communication runs entirely or almost entirely through technology: AI chatbots, rule-based bots and phone menus. It would apply to every business operating in New York City, even if it is based elsewhere.

Four obligations sit at the center:

- Businesses disclose that an automated system is answering. Under the draft, this happens at the start of the conversation and in the language it is held in.

- There is an easy way to reach a human.

- Anyone who promises customer service delivers real help.

- Businesses cannot argue that the technology caused the violation.

Under the draft, a business acts unfairly if it repeatedly and without good reason sends customers into menus that don’t solve their problem, or responds with irrelevant, unhelpful or wrong answers. In the US, this is known as the customer service “doom loop”.

Fines rise with each violation: $525 for the first, $1,050 for the second and $3,500 from the third onwards. Affected customers would also be compensated. Businesses would have to keep records, for example on how long conversations lasted, whether a human was offered and whether anyone was transferred.

For now, it is only a proposal. The public hearing and the end of the comment period both fall on November 19, 2026. The agency will then review the submissions. Once adopted, businesses are to get a transition period.

## What is surprising about the rule?

California, Maine, New Jersey and Utah already require bots to identify themselves as bots. California has also passed a law requiring very large companies to offer a path to human help. New York goes further: the rule would apply to all businesses and describes the exit route in detail.

The draft works with a rebuttable presumption. A business that runs a bot without a route to a human is presumed to be acting unfairly. It can rebut this with a single-step exit, such as typing or saying “agent” or pressing 0. If customers have to wait, the system states the estimated wait time. If no one is available, there has to be an alternative: service hours, a phone number, live chat, or a callback, email or text message with a time frame for the reply.

It reads like a checklist from a conversation design workshop. Now it is in a draft regulation, with fines attached.

DCWP Commissioner Samuel Levine also wants New York to set an example. He told local news outlet [Gothamist](https://gothamist.com/news/nyc-proposes-crackdown-down-on-ai-customer-service-chatbots) that he believes every jurisdiction should follow suit.

## Why do bots send customers in circles?

The agency backs the rule with figures from Gartner. According to them, chatbots resolve only 17 percent of billing disputes and 58 percent of issues around returns and cancellations. The Gartner figures date from 2023.

We see two causes that have little to do with the language model.

The first is measurement. Many bots are judged by how many conversations they close without a human. Every handover then counts as a failure, and teams build the bot to hold on to customers. For customers, that is exactly the loop.

The second is missing system access. A bot that cannot trigger a refund or look up an invoice can only explain and stall when it gets these requests. It goes in circles because it isn’t allowed to do anything. We’ve written about how users prefer quick answers to long conversations in [Nobody wants to talk to your chatbot](https://www.lautmaler.de/blog/no-one-wants-to-talk-to-your-chatbot).

Then there are wrong answers. Even a well-built bot gets things wrong sometimes. Why that won’t change is explained in [Why language models will always hallucinate](https://www.lautmaler.de/blog/ai-hallucinations). All the more reason to have an exit customers can always find.

## Does anything similar apply in Germany and the EU?

There is no comparable rule in Germany or the EU. Since August 2, 2026, however, [Article 50(1) of the AI Act](https://ai-act-service-desk.ec.europa.eu/en/ai-act/article-50) has applied. Providers must design AI systems that interact directly with people so that those people know they are dealing with an AI, unless this is obvious anyway. The notice must come at the latest at the first interaction. Article 50 does not include a right to a human.

On liability, a German court has already ruled. The Higher Regional Court of Hamm (OLG Hamm) decided that businesses are liable for false statements made by their AI chatbot. What the ruling says in detail is covered in “[The chatbot did it](https://www.lautmaler.de/blog/chatbot-liability-hamm-ruling)” no longer works. Here, too, technology is no excuse.

If you have customers in New York, keep an eye on the rule. It is meant to apply to every business operating there.

## How do you design a chatbot handover to a human?

Regardless of the law, it is worth asking whether your bot would pass the New York test. Five points:

1. **One-step exit.** One word in chat, one word or one key on the phone. It has to work in every state of the conversation, including halfway through a form.

2. **Be honest about wait times.** Customers being transferred learn how long it will take. If no one is available, the bot gives service hours or offers a callback with a time frame.

3. **Pass on the history.** The person on your team sees what has already been asked and tried. Nobody has to explain their issue twice.

4. **Detect loops.** If someone asks the same question a second time or types “human”, the bot offers the handover on its own.

5. **Measure differently.** A good handover counts as a success. The number that matters is the share of issues resolved, whether by the bot or by your team.

The same goes for voicebots. On the phone, a loop is even more draining, because nobody can scroll down to find a contact link.

## Conclusion: The exit is part of the bot

New York is putting into a draft regulation what good conversation design already requires. A bot can take on a lot, as long as customers know who they are talking to and can reach a human in one step. Businesses that hide the exit to save costs may soon face fines in New York. Everywhere else, they pay with their customers’ trust.

We have been building chatbots and voicebots in Berlin since 2007. We connect them to your actual systems so they can resolve issues, and we design the handover to your team from the start. How many steps does it take your customers to reach a human? Talk to us [→ EN-URL von /kontakt]

*This article is a general overview and not legal advice. Please consult a lawyer to assess your specific case.*

## FAQ: Chatbot handover to a human

### **What is a doom loop in customer service?**

It is the US term for an endless loop with automated systems. A chatbot or phone menu keeps sending customers to menus or answers that don’t solve their problem, and there is no way to reach a human. New York City wants to treat this as an unfair business practice.

### **Does a chatbot in Germany have to transfer customers to a human?**

There is no general legal obligation to do so in Germany. Since August 2, 2026, however, AI systems must make clear that people are talking to an AI (Article 50 of the AI Act). Businesses are liable for false statements made by their chatbot, as the Higher Regional Court of Hamm has ruled.

### **What should a chatbot handover to a human look like?**

A single step has to be enough, such as one word in chat or one key on the phone. The bot states the wait time and passes the conversation history to the team. If no one is available, it offers service hours or a callback with a time frame.

### **When does New York’s chatbot rule take effect?**

Not yet. The consumer protection agency proposed it on October 20, 2026, and the hearing and comment period end on November 19, 2026. The agency can then amend and adopt the rule, with a transition period for businesses.

### **Does the rule also cover voicebots and phone menus?**

Yes. The draft covers all customer service systems where communication runs entirely or almost entirely through technology. That includes AI chatbots, rule-based bots and phone menus.

Quote

/ FAQ

The key points of this article at a glance. Your question isn't here? **Then just chat with our bot!**

[Ask ChatLord](https://www.lautmaler.de/en/contact)

## You have a Project Idea?

Let's talk and see what we can do together.

[Get in Touch](https://www.lautmaler.de/en/contact)
