NYC AI chatbot doom-loop rule: $525 fines, human off-ramp

If your support bot can't answer a question and won't let the customer reach a person, New York City wants to call that a deceptive trade practice. On October 20, the city's Department of Consumer and Worker Protection proposed an AI chatbot doom-loop rule for automated customer service: disclose the bot, offer an easy route to a human, and actually resolve the problem you claim to resolve. Fines would start at $525, plus restitution. If you sell to New Yorkers through a chatbot or a phone tree, this one lands on you, not your vendor.

What actually happened

Per DCWP's announcement, the proposed rule was published in the City Record on October 20 and covers AI chatbots and telephone voice response systems. It would:

• Codify doom loops as deceptive and unconscionable: advertising customer service, then repeatedly failing to address inquiries or answering with irrelevant or wrong information.
• Require disclosure when a customer is talking to an automated system.
• Require an easy way to reach a human representative.
• Block the "the AI did it" defense. Businesses can't avoid responsibility when their AI use produces unlawful conduct.

Violators would owe restitution to harmed consumers and civil penalties starting at $525. A 30-day comment period with a public hearing comes first, then DCWP reviews submissions before adopting the rule. No effective date is set yet. The release also points to a September joint policy statement that says the City Consumer Protection Law already covers what chatbots and other agentic systems tell customers.

Why the chatbot rule matters for your business

"Talk to a human" is now a feature you ship, not a favor. Most doom loops are not malice. They are a bot with no escalation path because nobody wrote one. Add a handoff trigger: the customer asks for a person, the bot fails twice on the same intent, or the topic is billing, cancellation, or a refund. Route it to a queue someone watches.

Your bot's answers are your answers. The city already treats chatbot statements as business statements. If the bot invents a return policy, you own it. Ground replies in your actual policy docs and log every conversation so you can prove what was said.

One label covers disclosure. "You're chatting with our AI assistant" in the first message costs nothing. Put it in the greeting and in your IVR script now.

Measure resolution, not deflection. DCWP's release cites Gartner figures of chatbot resolution ranging from 17% on billing disputes to 58% on returns and cancellations. If your dashboard celebrates "tickets deflected," you are measuring the exact behavior this rule targets.

Sources: NYC Department of Consumer and Worker Protection.