AI calls and the law: consent, AI disclosure, the EU AI Act and country rules—FAQ

In many countries, yes. In the US, the regulator confirmed in February 2024 that calls using an AI voice fall under the ban on automated calls without prior consent (Hunton); getting it wrong costs $500–1,500 per call (ActiveProspect). In the UK, automated marketing calls always require consent, including calls to businesses (ICO). In Canada, automated calls need express consent (CRTC). In Germany, advertising via an “automated calling machine” requires express consent even between businesses, with fines up to €300,000 (Famulor, secondary source).

In the EU, yes. Since August 2, 2026, the EU AI Act's transparency requirements apply: people must know right away that they're talking to AI. Fines go up to €15 million or 3% of global turnover (Goodwin). The US has no general law, but some states are adding their own disclosure requirements. Our recommendation: in the EU and UK, disclose in the first sentence; elsewhere, at minimum give an honest answer to a direct “are you a robot?”

The EU AI Act's transparency obligations took effect for systems that interact with people. For voice agents, that means disclosing AI at the start of every call into the EU—no exceptions, and the client can't switch that line off.

Yes, but you have to say so. In Germany, recording without consent is a criminal offense; in the UAE, you must announce recording at the start of the call (Legal500). You can't publish recordings of real clients without their consent—which is why the samples on our site use synthetic voices.

Since August 2024: calls only from registered numbers, 9:00 a.m. to 6:00 p.m., never to numbers on the do-not-call registry, fines from AED 50,000. 159 companies have already been fined for violations (Legal500, Gulf News). Public sources don't say outright whether these rules cover calls to businesses from abroad. The safe approach: message first, call after a reply.

In France, since August 11, 2026, commercial calls require consent (Connexion France); sources disagree on whether that covers calls to businesses. In India, marketing calls are allowed only from special number series (per the regulator TRAI, 2025). South Africa has run a national opt-out registry since 2026 (gov.za). For Brazil and Saudi Arabia, we recommend the same cautious approach: message first, call businesses during business hours and only after checking the rules.

Where there's no direct ban on automated calls to businesses—for example, in many Latin American countries. But rules are tightening there too, so check them before every launch.

No. A business can only call someone on WhatsApp who has accepted a call request, and it can send no more than one request a day and two a week (Infobip). Meta treats calls without consent as a violation of its rules. Details in WhatsApp vs. mobile.

Contracts usually make the client responsible for the legality of its list and for having consents. But regulators penalize both the client and the contractor: in Canada, companies were fined because their contractor broke do-not-call registry rules (insurance-portal.ca). So the vendor checks the list too and has the right to turn down a questionable one.

No. In August 2025, the US Federal Trade Commission sued Air.ai, which sold AI calling with promises of fast payback (FTC). What you can honestly guarantee is how the system works—number of calls, recordings, reports—not revenue.

Countries under comprehensive sanctions, and destinations where our voice and telephony providers don't carry traffic. Before launch, every list is screened against US, UN, EU and UK sanctions lists.