Can Your AI Agent Legally Call That Number? TCPA Basics

I'm not a lawyer and this isn't legal advice. It's the base I've learned for running outbound campaigns without getting into stupid trouble.

If you're a business owner thinking about having a voice agent make outbound calls, this is what you need to understand first. These laws aren't theory. In February 2026, Mortgage One Funding got hit with two class action lawsuits in 15 days. They ran AI cold calls about refinancing, with no consent, to numbers on the Do Not Call list.

Your AI agent counts as a robocall

In February 2024 the FCC unanimously ruled that an AI generated voice is an artificial voice under the TCPA. That means your agent is regulated exactly like a prerecorded message, under rules that go back to 1991. Doesn't matter how conversational it sounds.

A big reason why: right before the New Hampshire primary in January 2024, about 5,000 voters got a call from a Joe Biden voice clone telling them to stay home. The FCC proposed a $6 million fine against the organizer, and the carrier that let the traffic through paid $1 million.

Fines are per call. Not per campaign, not per complaint. Per call. That's $500 each, and up to $1,500 if a court decides you did it willfully or knowingly. 100 bad calls at $500 is $50,000. ViSalus made about 1.8 million illegal calls and got hit with $925 million. Dish Network made 51,000 and got $61 million.

Who you can call: consumer or business

Every outbound list goes through one question first. Whose number is this? A consumer's personal phone, or a business line?

If it's a consumer and the call is marketing, meaning you're selling something, offering a promotion, or setting a sales appointment, you need prior express written consent. That's a signed agreement naming your company and that phone number.

If it's a consumer and the call is purely informational, like an appointment reminder with no upsell, the bar is lower. Prior express consent is enough. But watch it. If the agent drifts into "by the way, we're running a discount on such and such," that reminder just became a marketing call and the written consent rules apply.

Business to business is way simpler than people think. All the noise about AI agents needing written consent to call out is about consumers. Calling a verified business landline without written consent is fine, even for a marketing call. The catch: plenty of contractors, spas, HVAC and roofing companies use a personal cell as the business number. Cross those off, because you're right back in consumer territory.

What real consent looks like

A checkbox alone doesn't cut it. The language matters. Mine reads along these lines: I agree that Oceanside AI Solutions may contact me at the number provided using an automatic dialing system or an artificial or prerecorded voice, and text messages, for marketing purposes. I understand consent is not required and that I can opt out at any time by replying STOP.

If it doesn't mention an artificial or prerecorded voice, it does NOT cover your AI agent. And texting permission through A2P is not the same as permission for AI voice calls. Two different things.

You also need proof. The person suing never has to prove they didn't consent. You have to prove they did. So keep the timestamp, the checkbox, their phone number, and the exact disclosure text. The IP address isn't technically required right now, but I'd collect it anyway.

Old leads come up a lot. A business wants to point an agent at a pile of old leads to bring them back. If those leads never gave proper written consent, you can't just call them to ask for it. Email them, explain clearly what you're asking, send them to a landing page with an unchecked opt in box, and only call the people who opt in. Painful, but that's how it works.

What your agent has to do on every call

Three things on every artificial voice call. Identify who's calling. Give a callback number. And keep an internal do not call list, so when someone says stop, the agent records it somewhere you can check later, even a Google Sheet.

Does it have to announce it's an AI up front? At the federal level, no. California's AB 2905, effective January 1, 2025, does require the agent to say it's an AI at the very start of the call. My rule of thumb everywhere: if someone asks, the agent answers honestly. Yes, I'm an AI.

States also stack their own rules on top of the federal ones. Colorado, for example, has its own do not call registry separate from the national list. Check every state you plan to call before launch. I go through a preflight checklist on all of this before any campaign, and I'd want anyone running calls for your business to do the same.