
Are AI calls legal?
Key Facts
- AI calls are legal but classified as robocalls, facing $500 to $1,500 statutory damages per violation according to legal analysis.
- Outsourcing AI calls does not transfer liability; the brand remains legally responsible for every violation per recent litigation trends.
- Established business relationships do not exempt AI calls from consent requirements, unlike live human agents according to compliance experts.
- Non-compliant 10,000-call campaigns create potential exposure of $5 million to $15 million in damages based on statutory limits.
- Opt-out requests must be honored within 10 days, with automated options delivered within two seconds per FCC guidelines.
- Texas mandates AI disclosure within the first 30 seconds of a call, setting a strict state standard according to state laws.
- Eleven states, including California and Florida, require all-party consent for call recordings per legal reviews.
The Legal Reality: AI Calls Are Legal, But Regulated Like Robocalls
On February 8, 2024, the FCC settled the question that had been hanging over the voice AI industry: AI-generated voices count as "artificial or prerecorded voice" under the TCPA, no matter how convincingly human they sound. The ruling, covered in NCLC's roundup of recent TCPA developments, means an AI call can't sidestep robocall rules by claiming to be the equivalent of a live agent. The statute doesn't allow that carve-out, and the FCC said so directly.
So AI calls are legal. What determines whether your calls are legal comes down to three things: consent, disclosure, and opt-out handling.
Consent is tiered, and it's always required. Marketing calls need prior express written consent; informational calls need prior express consent, which can be oral. As one compliance analysis puts it, for AI voice calls the only question is which level of consent applies. The stakes for getting this wrong are steep: statutory damages run $500 to $1,500 per call, with no aggregate cap. A 10,000-call campaign that misses the mark carries $5M to $15M in potential exposure. And an established business relationship does not exempt AI calls the way it exempts your live sales reps, a counterintuitive trap for anyone planning AI follow-up to existing customers.
Disclosure and opt-out mechanics round out the requirements:
- Identify the responsible entity at the start of the call and provide a callback number.
- Deliver an automated opt-out within two seconds of the initial message, and honor a revocation within 10 days.
- Recognize natural-language opt-outs like "stop," since consumers can revoke consent through any reasonable means.
- Respect calling hours: nothing before 8 a.m. or after 9 p.m. local time.
One more thing worth knowing: liability follows the brand, not the vendor. Outsourcing doesn't transfer compliance risk.
Here's the reassuring part for most small businesses. Outbound campaigns carry the bulk of this risk. Inbound AI receptionists, the calls that come to you, sit in a materially lower-risk category with simpler disclosure requirements. If you're weighing a done-for-you build like those from Agents by AIQ, the right questions are about audit-ready records, real-time DNC scrubbing, and consent that's queryable at call time. If consent can't be confirmed, the call shouldn't go out.
The Traps That Catch Business Owners Off Guard
Most business owners assume that existing customer relationships cover them legally. They are wrong. The law cares about the voice, not the history.
A live SDR can dial a 16-month-old customer on the Do Not Call list under the Established Business Relationship exemption. An AI agent cannot dial that same person without separate consent. This distinction trips up many teams who assume their CRM data is enough. It is not. The FCC rules tie compliance to the technology used, not the relationship held.
You might think outsourcing the calls to a vendor shifts the legal risk. It does not. Liability follows the brand, not the tool. In cases like Lamb v. Mortgage One Funding, courts target the entity on whose behalf the calls were made. Outsourcing does not transfer compliance risk to the provider. You remain on the hook for every violation.
State laws add another layer of complexity. Federal rules set the baseline, but state statutes stack on top. Texas requires AI disclosure within the first 30 seconds of a call. Eleven states, including California and Florida, mandate all-party consent for recording. If you operate nationally, you must design for the strictest standard to avoid a patchwork of violations.
The financial stakes are concrete and severe. Statutory damages range from $500 to $1,500 per call with no aggregate cap. A single non-compliant campaign of 10,000 calls creates potential exposure of $5 million to $15 million. Recent class settlements illustrate the scale of this risk:
- QuoteWizard settled for $19 million in a TCPA class action
- Gen Digital paid $9.95 million to resolve similar claims
- Hy Cite agreed to a $4.75 million settlement
These figures are not hypothetical. They represent real businesses that failed to verify consent or disclose AI usage properly. The audit trails generated by AI campaigns can be used against you. Plaintiffs' attorneys use these logs to prove patterns across entire classes.
Compliance must be built into the infrastructure from day one. This means queryable consent records, real-time DNC scrubbing, and immediate opt-out handling. Agents by AIQ designs these systems to ensure that if consent cannot be confirmed, the call simply does not go out. This approach protects your brand from the kind of liability that follows the caller, not the vendor.
The regulatory landscape is shifting, with new rules pending and others vacated. Staying compliant requires more than a policy document. It requires a system that enforces the rules automatically. If you are deploying AI voice agents, verify that your vendor provides audit-ready records and robust disclosure scripts. The cost of compliance is far lower than the cost of a settlement.
What Compliant AI Calling Actually Looks Like
AI-generated calls are legal in the United States. The harder question is whether a given call can prove it was compliant, and that proof has to live in the system's infrastructure rather than in a policy document.
Three pillars carry most of the weight: tiered consent, disclosure at call start, and a working opt-out. Get them right and the rest is configuration. Get them wrong and statutory damages run $500 to $1,500 per call with no aggregate cap, according to Henson Legal's AI voice compliance analysis.
Consent is tiered by purpose. Marketing calls require Prior Express Written Consent; informational calls require Prior Express Consent, which oral consent satisfies. Consent is always required for an AI voice call, so the only real question is which tier applies. That written-consent floor is unsettled: Henson Legal notes a recent Fifth Circuit ruling suggesting the statute requires only prior express consent, so written consent stays the safer assumption for anything promotional.
Disclosure comes at call start, identifying the responsible entity and giving a callback number. State rules already push that way: Texas requires AI disclosure within 30 seconds, and the FCC has a pending proposal that would mandate in-call AI disclosure. Because the rules keep moving, announcing the AI disclosure and callback number before any sales content is the safe default.
The opt-out runs on a clock. The automated opt-out must be delivered within two seconds of the initial message, and a revocation request must be honored within 10 days. Consumers can revoke by any reasonable means, including saying "stop," which is why natural-language opt-out recognition matters. Answered calls also face a 3% abandonment cap, where abandonment means no response within two seconds of the consumer's greeting.
Here is where most policy documents miss. Compliance has to be engineered into the stack through a small set of components:
- Call logging with timestamps and outcomes
- Consent records queryable at call time
- Real-time do-not-call scrubbing before a dial
- Records producible on demand
The operating rule is simple: if consent cannot be confirmed, the call does not go out. The audit trail cuts both ways. Because AI campaigns are logged and scriptable, plaintiffs' lawyers can prove a pattern across an entire class from your own records. The trail that protects a compliant operator convicts a non-compliant one.
This is also the strongest argument for AI agents over human callers, and it is a vendor framing rather than a legal fact. AI agents are programmable, consistent, and auditable, so a disclosure script, an opt-out handler, and a consent check behave the same on the ten-thousandth call as on the first.
Agents by AIQ builds and runs done-for-you AI agents for small and mid-size businesses, including AI receptionists that answer on a real phone number. Inbound answering carries materially lower compliance risk than outbound campaigns. Weighing AI follow-up? Ask any vendor for audit-ready records, do-not-call scrubbing, and disclosure scripts, then book a scoping call to map your consent and state exposure. This is general compliance information, not legal advice.
How Agents by AIQ Builds Compliance Into the Agent
Knowing the rules is one thing. Building an agent that actually follows them, call after call, is where most businesses get into trouble, because as compliance practitioners put it, compliance has to live in infrastructure, not in a policy document someone wrote once and filed away.
That's the philosophy behind how Agents by AIQ approaches every build. The legal requirements we've covered in this article, consent tiers, calling-hour windows, opt-out handling, and record retention, aren't bolted on afterward. They're designed into the agent from the first sketch.
Consent is the clearest example. Because the FCC treats AI-generated voices as artificial voices under the TCPA, consent is always required for AI voice calls; the only question is which tier applies, according to legal analysis of the 2024 ruling. So every build logs consent records with timestamps and links them to the contact, so the agent can confirm consent at call time. If consent can't be confirmed, the call doesn't go out.
Here's what a done-for-you build is designed around:
- Consent records with timestamps, queryable at the moment a call is placed, so you can prove provenance years later. That matters because the TCPA statute of limitations runs four years, and defense counsel recommends keeping records for seven.
- Calling-hour windows enforced in code, keeping outbound activity inside the federal safe zone of 8 a.m. to 9 p.m. local time for the person being called.
- Natural-language opt-out recognition, so a caller who says "stop calling me" is honored on the spot. Consumers can revoke consent through any reasonable means, and missed revocations are a top lawsuit driver, with penalties running $500 to $1,500 per call with no cap.
- Audit-ready call logs, including transcripts and outcomes, producible on demand. The same audit trail that convicts a non-compliant operator protects a compliant one.
One point worth underlining: liability follows the brand, not the vendor. Outsourcing your calling doesn't transfer compliance risk, which is exactly why you should own your records. With an Agents by AIQ build, you do. The client owns everything: the consent data, the call logs, the agent configuration. If you ever change providers or face a records request, your evidence is yours, not trapped in someone else's platform.
A few honest caveats. These capabilities reflect general legal requirements that the build is designed around; they aren't verified legal claims about your specific situation, and no source we reviewed evaluates any particular vendor's compliance in practice. The rules are also still moving, with federal disclosure requirements pending and state laws like Texas's 30-second AI disclosure rule already in effect.
None of this is legal advice. Before launching any AI calling campaign, especially outbound marketing, confirm the specifics with qualified counsel who knows your state and your industry. If you'd like to explore how a done-for-you agent that answers your calls and follows up on leads could work within those requirements, book a scoping call and we'll talk through your setup.
Questions to Ask Before You Deploy Any AI Caller
Before you sign anything, put the vendor on the witness stand. Under the TCPA, you carry the liability even when a third party makes the calls on your behalf, and recent litigation shows courts targeting the brand, not just the vendor. With statutory damages of $500 to $1,500 per call and no aggregate cap, a bad answer here is expensive.
Start with consent. Ask exactly how the vendor captures it, where the record lives, and whether it's queryable at call time. The standard worth insisting on is simple: if consent cannot be confirmed, the call does not go out. And watch for the trap that catches most owners: the established business relationship exemption lets your live rep call a past customer on the DNC list, but it does not extend to AI voice calls. The voice is what the law cares about.
Then ask how Do Not Call scrubbing happens. It should be real-time, at the moment of dialing, not a weekly batch export. A list scrubbed on Monday is stale by Friday.
Third, get the disclosure script in writing. Several states now mandate it: Texas requires AI disclosure within 30 seconds, and California, Florida, Colorado, Illinois, and Utah have their own variants. A compliant call identifies the responsible entity at the start and provides a callback number.
Fourth, ask how recordings are handled. Eleven states, including California, Florida, Illinois, Pennsylvania, and Washington, require all-party consent to record. For a nationwide caller, the only scalable approach is to design as if all-party consent applies everywhere and announce recording at the outset of every call.
Finally, ask what happens when someone says "stop." Consumers can revoke consent by any reasonable means, including saying it out loud mid-call, and businesses have 10 days to honor a revocation. An AI agent that recognizes natural-language opt-outs addresses one of the top drivers of TCPA lawsuits. One that keeps talking after "stop calling" creates the record a plaintiff's attorney will use against you, because your own call logs prove the pattern across an entire campaign.
Here's the short version to take into any vendor conversation:
- How is consent captured, stored, and confirmed at call time?
- Does DNC scrubbing happen in real time, before every dial?
- What exact disclosure script does the agent open with, and does it meet state rules like Texas's 30-second requirement?
- How are recordings announced and stored for all-party consent states?
- How does the agent detect and act on a spoken opt-out, and how fast?
Compliance has to live in the infrastructure itself: call logs, consent linkage, real-time scrubbing, and records you can produce on demand. That's how we approach every agent we build at Agents by AIQ, designing the guardrails into the system rather than hoping the script holds. If you're evaluating an AI caller for your business, book a scoping call and we'll walk your use case through each of these requirements before a single number is dialed. This article is general information, not legal advice; confirm specifics with your attorney.
Frequently Asked Questions
Are AI-generated phone calls actually legal in the US?
Can I use an AI agent to call my existing customers without new consent?
What are the penalties if my AI calls violate TCPA rules?
If I outsource AI calling to a vendor, who is legally liable?
What disclosure and opt-out rules do AI calls have to follow?
Do state laws add extra requirements on top of federal AI call rules?
Making Compliance a Feature, Not a Risk
The legal landscape for AI calling is clear: the technology is permitted, but the liability is absolute. As we’ve seen, the FCC treats AI voices as artificial, meaning consent, disclosure, and opt-out handling are non-negotiable. Because liability follows the brand rather than the vendor, your protection lies in how your systems are engineered. A compliant AI agent isn’t just a tool; it’s an audit-ready infrastructure that logs consent in real-time and honors revocations instantly. This is where the operational advantage of a done-for-you build becomes critical. By designing these guardrails into the agent from the start, you ensure that if consent can’t be confirmed, the call simply doesn’t go out, protecting your business from the significant statutory penalties associated with non-compliance. The rules are moving, but the core requirement for safety remains consistent: robust, queryable records. If you are ready to deploy an AI receptionist or follow-up agent that handles these legal nuances automatically, book a scoping call with Agents by AIQ. We’ll map your specific consent and state exposure, ensuring your agent is built for compliance before a single number is dialed.