AI Voice in France 2026: ARCEP, Numéro Vert, and the August Cold-Calling Ban

France just made AI voice cold-calling almost impossible. As of 1 January 2026, ARCEP's MAN (Mécanisme d'Authentification des Numéros) is fully operational and blocks spoofed CLIs. From 11 August 2026, every commercial call to a French personal number requires prior explicit consent — opt-out is dead, opt-in only.

How telephony works there

ARCEP (Autorité de régulation des communications électroniques, des postes et de la distribution de la presse) administers the French numbering plan. Geographic numbers are 10 digits with five regional zones (01 Île-de-France, 02 NW, 03 NE, 04 SE, 05 SW). Mobile is 06/07. Non-geographic ranges include 0800-0805 (numéros verts, free to caller), 081x/082x/089x (numéros surtaxés, premium-rated), and 09 (NGN — non-geographic, untimed-local-rate).

Twilio offers French DIDs across the five geographic zones and numéros verts (0805 mostly). The Paris media edge keeps RTT under 30 ms. PSTN-IP migration (RTC switchover) is largely complete, with Orange having retired residential analogue lines through 2023-2025. CLI rules require a valid presentation number; from 1 January 2026, every call must carry a MAN authentication token at the SIP layer or it is blocked at the originating carrier.

Regulatory landscape

Three regimes apply, all tightening. First, ARCEP's numbering plan and the MAN: a Stir/Shaken-style authentication token must be attached to every call originating from or transiting France. Second, Bloctel: the historical opt-out register that since 2025-2026 has shifted toward opt-in for personal numbers. Third, the law of 30 June 2025 on telemarketing — effective 11 August 2026 — bans commercial calls to personal numbers without explicit prior consent. Soft opt-in via existing-customer relationships is no longer enough; the consent must be specific to the channel, recent, and revocable.

The CNIL enforces GDPR overlap: voice biometrics is sensitive data and requires explicit consent; recordings have a 6-month default retention unless extended for legitimate purpose. The DGCCRF can fine up to €375,000 per legal-entity violation for unsolicited commercial communications.

CallSphere implementation

Explore a live demo and compare current plans to find the right fit for your business.

Build and launch steps

flowchart LR
 A[Twilio FR subaccount + Paris edge] --> B[Provision zone DID + 0805]
 B --> C[Register CLI under RCS + MAN auth token]
 C --> D[Opt-in consent capture for personal numbers]
 D --> E[AI disclosure first sentence in French]
 E --> F[Bloctel scrub + DGCCRF audit log]

1. Create a Twilio French subaccount; pin media to Paris.
2. Provision a zone-matched geographic DID and a 0805 numéro vert.
3. Register CLI to the RCS-registered address; ensure MAN authentication tokens are emitted at SIP.
4. From August 2026, capture opt-in consent for any commercial call to a personal number; B2B remains under implied consent.
5. Configure the AI agent to disclose AI in the first sentence in French; log disclosure for CNIL/DGCCRF audits.
6. Run Bloctel scrub daily; honour CNIL ARCO rights; cap recording retention at 6 months unless justified.

FAQ

What changes on 11 August 2026? Every commercial call to a French personal number requires prior explicit consent. Bloctel-style opt-out is no longer the operative regime for personal numbers.

What is MAN? Mécanisme d'Authentification des Numéros, France's caller-ID authentication framework, fully operational from 1 January 2026. Spoofed or unauthenticated calls are blocked.

Are B2B calls also opt-in only? No. B2B remains permitted under implied consent grounded in the recipient's likely professional interest. Personal numbers are opt-in only.

Can I record a French call? Yes, with explicit opt-in. CNIL guidance is a 6-month default retention unless extended for a legitimate documented purpose.

Sources

• ARCEP: Numbering Plan press release
• ARCEP: Consumer protection (caller-ID authentication)
• Connexion France: New rules to combat phone spoofing
• Nixxis: Cold-calling rules under the 11 Aug 2026 law

Explore a live demo and compare current plans to find the right fit for your business.

AI Voice in France 2026: ARCEP, Numéro Vert, and the August Cold-Calling Ban: production view

AI Voice in France 2026: ARCEP, Numéro Vert, and the August Cold-Calling Ban is also a cost-per-conversation problem hiding in plain sight. Once you instrument tokens-in, tokens-out, tool calls, ASR seconds, and TTS seconds against booked-revenue per call, the right tradeoff between Realtime API and an async ASR + LLM + TTS pipeline becomes obvious — and it's almost never the same answer for healthcare as it is for salons.

Serving stack tradeoffs

The big fork is managed (OpenAI Realtime, ElevenLabs Conversational AI) versus self-hosted on GPUs you operate. Managed wins on cold-start, model freshness, and zero-ops; self-hosted wins on unit economics past a certain conversation volume and on data residency for regulated verticals. CallSphere runs hybrid: Realtime for live calls, self-hosted Whisper + a hosted LLM for async, both routed through a Go gateway that enforces per-tenant rate limits.

Latency budgets are non-negotiable on voice. End-to-end target is sub-800ms ASR-to-first-token and sub-1.4s first-audio-out; anything beyond that and turn-taking feels stilted. GPU residency in the same region as your TURN servers matters more than choosing a slightly bigger model.

Observability is the unglamorous backbone — every conversation produces logs, traces, sentiment scoring, and cost attribution piped to a per-tenant dashboard. HIPAA aligned isolation keeps healthcare traffic separated from salon traffic at the storage layer, not just the API.

FAQ

How does this apply to a CallSphere pilot specifically? Setup runs 24 hours, the trial is 7 days with no credit card, and pricing tiers are $49, $99, and $149 — so a vertical-specific pilot is a same-week decision, not a quarterly project. For a topic like "AI Voice in France 2026: ARCEP, Numéro Vert, and the August Cold-Calling Ban", that means you're not starting from scratch — you're configuring an agent template that's already been hardened across thousands of conversations.

What does the typical first-week implementation look like? Day one is integration mapping (scheduler, CRM, messaging) and prompt tuning against your top 20 real call transcripts. Day two through five is shadow-mode running, where the agent transcribes and recommends but a human still answers, so you can compare side-by-side. Go-live is the moment your eval pass-rate clears your internal bar.

Where does this break down at scale? The honest answer: it scales until your tool catalog gets stale. The agent is only as good as the integrations it can actually call, so the operational discipline is keeping schemas, webhooks, and fallback paths green. The platform handles the rest — observability, retries, multi-region routing — without your team owning the GPU layer.

Talk to us

Explore a live demo and compare current plans to find the right fit for your business.