TCPA Compliance for AI-Powered Outbound Calls
TCPA — the Telephone Consumer Protection Act — is the federal law that governs automated and pre-recorded outbound calls in the United States. AI-generated voice calls fall squarely under TCPA's stricter rules for "artificial or prerecorded voice" messages.
TCPA — the Telephone Consumer Protection Act — is the federal law that governs automated and pre-recorded outbound calls in the United States. AI-generated voice calls fall squarely under TCPA's stricter rules for "artificial or prerecorded voice" messages. Every voice AI operator doing outbound has to understand and comply with TCPA, or face class-action litigation that can run to tens of millions of dollars. This piece covers the practical compliance realities — what the law requires, how to operationalize it, and the specific issues AI-driven outbound raises.
TL;DR
- TCPA regulates automated and AI voice outbound calls in the US.
- Prior express consent (PEC) required for non-marketing; prior express written consent (PEWC) for marketing.
- AI voice = "artificial or prerecorded" for TCPA — stricter rules apply than live human calling.
- Penalties: $500–$1,500 per violation, stackable.
- Operationalize: consent capture, DNC compliance, opt-out enforcement, audit trails.
What TCPA requires
Consent. For most AI outbound calls to mobile numbers, you need prior consent. For marketing to mobile via AI voice: written consent (PEWC).
Identification. The call must clearly identify:
- The calling business.
- The nature of the call.
Opt-out. Every AI call must include an immediate, automated opt-out mechanism.
Quiet hours. No calls before 8 AM or after 9 PM recipient's local time.
DNC compliance. Respect the federal Do-Not-Call registry and state equivalents.
Disclaimer: not legal advice. Consult counsel.
Consent types
Prior Express Consent (PEC). Non-marketing calls (informational, transactional). Can be oral or written.
Prior Express Written Consent (PEWC). Marketing calls. Must be:
- Written.
- Clear and conspicuous (not buried in fine print).
- Specific (mention that AI calls may occur).
- Revocable.
- Dated and identifiable to a person.
For AI outbound marketing, PEWC is the bar.
What counts as "written"
- Checkbox on a web form with clear language.
- E-signature on an agreement.
- SMS double opt-in ("Reply YES to confirm").
- Voice consent captured and recorded.
Not:
- Implied consent from providing a phone number.
- General terms of service acceptance.
- Buried-in-fine-print language.
The AI voice regulation
FCC rulings since 2024 have clarified:
- AI-generated voice in outbound calls = "artificial or prerecorded voice" under TCPA.
- The stricter rules for artificial voice apply regardless of how realistic the AI sounds.
- No carve-out for "sounds human." The rule is about how the call is generated, not how it sounds.
Practical consequence: AI voice outbound is held to the higher consent standard (PEWC for marketing).
DNC registry
Federal Do-Not-Call:
- Numbers on the registry should not receive marketing calls.
- Refresh your DNC scrub at least every 31 days.
- State DNC lists may be stricter.
Exemptions:
- Existing business relationship (with some limits).
- Consented callers.
Penalties
- $500 per negligent violation.
- $1,500 per willful violation.
- Stackable: each call is a separate violation.
- Class actions have produced $50M+ settlements.
Operationalizing compliance
1. Consent capture. At signup, purchase, event registration, etc. Log the exact consent text, timestamp, IP, method.
2. Consent storage. Queryable database. For any call, you can prove: "Yes, we have consent, dated X, via Y method."
3. Consent verification before calling. Pre-flight check on every outbound: do we have current consent?
4. DNC scrub. Daily or per-campaign scrub against federal + state DNC.
5. Opt-out enforcement. Caller says "don't call me again" → immediately added to internal suppression list. No exceptions.
6. Time zone enforcement. Area code + known time zone → calls only during 8 AM–9 PM recipient local.
7. Call logging. Every dial logged with metadata: number, time, outcome, consent status, campaign.
8. Audit. Random sample of calls regularly reviewed for compliance.
The disclosure
Every outbound AI call should open with:
- Business identification. "Hi, this is Acme's..."
- AI disclosure. "...AI assistant calling..."
- Purpose. "...about your account."
- Permission ask. "Is now a good time?"
No fine print. Upfront.
Opt-out keywords and phrases
Honor:
- "Stop"
- "Don't call me"
- "Remove me"
- "Put me on your don't-call list"
- "Opt out"
Immediately. Confirm verbally. File in suppression.
Ringless voicemail
Rules still evolving. Currently treated similarly to AI voice calls — TCPA applies. Don't rely on "it's not a real call" argument.
SMS TCPA
SMS has its own TCPA application:
- Similar consent rules.
- STOP must be universal opt-out.
- A2P 10DLC is the carrier-side compliance layer.
- Marketing SMS requires PEWC.
See A2P 10DLC explained for voice agent builders.
Common mistakes
"We got their phone number from [source]." Having a number isn't consent. Need explicit PEC/PEWC.
"They're an existing customer." EBR has limits. For marketing via AI voice, still need specific consent.
"We only called once." One call without consent = one violation = $500+.
"It was an informational call." If it advertises products or services, it's marketing. Intent doesn't override content.
"They didn't complain." Doesn't matter. A third party can file a TCPA class action even if your specific callers didn't.
What's allowed without PEWC
Legitimately transactional non-marketing:
- Appointment reminders you asked for.
- Delivery notifications.
- Account alerts (fraud, payment).
- Service outages.
- Emergency notifications.
These can go with PEC, not PEWC. But "emergency" has tight interpretation — don't stretch.
Revocation
Consent can be revoked any time:
- Via voice ("stop calling me").
- Via SMS ("STOP").
- Via email.
- Via any reasonable means.
Honor. Document. Add to suppression. Propagate across systems.
State layers
Several states have stricter rules:
- California. Constrains robocalls further.
- Florida (FTSA). State-level TCPA equivalent, sometimes broader.
- New York, Washington, others. Various.
Operating in multiple states means complying with the strictest applicable law.
Class-action landscape
Plaintiff-side TCPA firms actively monitor. Common patterns:
- Test calls placed to numbers on DNC registries.
- Monitoring for consent mismatches.
- Aggregating multiple violations into class actions.
Don't be a target. Compliance from day one.
Internal controls
- Compliance officer specifically responsible for TCPA.
- Regular audits (quarterly minimum).
- Vendor diligence — your voice AI vendor's TCPA posture.
- Training for anyone who configures outbound campaigns.
- Legal review of campaign content.
Related reading
- How to Build a Compliant Outbound Voice Agent in 30 Days
- Caller ID and Trust: Why Numbers Get Marked as Spam
- Outbound AI Calling in 2026: A Practical Playbook
- Voice Agents for Loan Servicing and Collections
- Outbound for B2B: Pipeline, Renewals, and Win-Backs
FAQ
What if we bought a lead list? Purchased consent doesn't transfer automatically. Verify each lead has valid consent for AI voice calls.
Can we call B2B numbers without consent? Landlines to businesses are less regulated. Mobile numbers are still subject to TCPA regardless of business/personal.
What if the caller is an AI researcher? Consent still needed. Doesn't matter who or why.
What about international calls to US numbers? TCPA still applies. Origin doesn't matter.
How do we defend against TCPA litigation? Paper trail. Strong consent records, DNC compliance, suppression, and documented ops.

Tyler Weitzman is co-founder and Head of AI at Speechify. He has spent the past decade building the speech-synthesis stack that powers millions of users. Tyler writes about the engineering of real-time conversational systems — text-to-speech, speech recognition, latency budgets, model serving, and the architectural choices that separate prototypes from production-grade voice agents.
More from Tyler Weitzman
View all →Open-Source vs Proprietary Voice Agent Stacks
The open-source voice AI stack in 2026 is genuinely good. Whisper and its derivatives handle STT. Open-weight LLMs like Llama 3/4, Qwen, Mistral handle the reasoning. Open-source TTS (XTTS, StyleTTS, Orpheus-class) handles output.
Build vs Buy: When to Build Your Own Voice Agent
Build-vs-buy for voice agents in 2026 is a different conversation than it was two years ago. Then, the open-source stack was rough and most serious deployments ended up building.
Voice Agents for Developer Support
Developer support is a strange category. Developers don't generally want to call anyone. They want Stack Overflow, they want clear docs, they want an LLM that can read their code.
Related reading
How to Build a Compliant Outbound Voice Agent in 30 Days
Getting an outbound AI voice agent live in 30 days sounds ambitious — but it's achievable for focused deployments. The critical path is compliance setup (TCPA, A2P 10DLC, number verification), not technology. The voice AI itself can be configured in a week.
Caller ID and Trust: Why Numbers Get Marked as Spam
You deploy an outbound voice AI campaign. First week goes great. Second week, answer rates drop 40%. Third week, your phone numbers start showing up as "Scam Likely" on caller ID. What happened?
A2P 10DLC Explained for Voice Agent Builders
If your voice agent sends SMS from a standard 10-digit US phone number, A2P 10DLC compliance is part of your stack — whether you know it or not.
Voice AI, twice a month.
Get the best of the SIMBA resources hub — new articles, trend notes, and operator guides. No spam.
