Two Compliance Deadlines Hit B2B Outbound This August: A Playbook

Two dates, nine days apart

I keep seeing these two rules mashed into one deadline in other outbound content, and they are not the same law, do not come from the same body, and do not apply to the same thing. Splitting them out matters because the fix for one is not the fix for the other.

The first is EU-wide: Article 50 of the EU AI Act, the transparency obligation, applies from August 2, 2026, per the law firm summary from McCann FitzGerald and the EU AI Act's own Article 50 reference. It covers any AI system built to interact directly with a person, not just chatbots on a pricing page.

The second is French: the telemarketing reform known as the Verzelen Law reaches full effect on August 11, 2026, ending the Bloctel opt-out registry and moving consumer telemarketing to an opt-in model, according to reporting from Nixxis and coverage from Connexion France on the law's passage.

What the EU AI Act's Article 50 actually requires

Article 50 is narrower than "label all your AI." For outbound teams, the part that matters is the interaction rule: a system built to talk directly with a person, a chatbot, a voice agent, an AI SDR that replies to prospects in a live chat, has to make it clear the person is talking to an AI, before or at the very start of the interaction. McCann FitzGerald's summary is specific that the duty does not apply "where the AI nature of the interaction is obvious to a reasonably well-informed, observant and circumspect person," so a clearly labeled bot widget is fine without an extra disclaimer, but a voice agent designed to sound like a rep is exactly the case the rule targets.

There is a second, separate piece: deployers who generate or alter synthetic audio, image, or video have to mark that output so it can be detected as artificial. That part has its own grandfather clause, applying from December 2, 2026 instead of August 2 for systems already on the market, which is the source of some of the date confusion floating around right now. The person-facing disclosure duty is the one on the August 2 clock.

Penalties are not small. Non-compliance can bring administrative fines "of up to €15 million or, for undertakings, up to 3% of total worldwide annual turnover, whichever is higher," per the same summary.

Why this covers more than a website widget

Most compliance write-ups I have read about Article 50 default to the chatbot example because it is the easiest to picture. Outbound teams should read the rule wider than that. If you run an AI SDR that places calls, leaves voicemails, or replies to inbound email threads in a way meant to pass for a human rep, that is the interaction Article 50 is written for, not a website FAQ bot. The obligation sits with anyone building or deploying a system meant to interact directly with people, which puts AI voice and AI email reply tooling for outbound squarely inside scope, not on the edge of it.

What changes in France on August 11

The Verzelen Law, named after the senator who introduced it, ends Bloctel, the do-not-call registry consumers had to opt into if they wanted to stop receiving calls. From August 11, 2026, French consumer telemarketing flips the default: a company cannot call a consumer commercially without prior, explicit consent. That is the headline most coverage leads with, and it is accurate for consumer calls.

It is not accurate for B2B. The scope of the reform is consumer telemarketing, meaning calls to a person acting outside a professional capacity. It does not touch B2B prospecting conducted under GDPR's legitimate interest basis, which is the framework most cold calling into French companies already runs on.

The B2B exemption, and why it is narrower than it sounds

"B2B is exempt" is true and also the sentence that gets outbound teams into trouble, because the exemption is about purpose, not about which phone number you dialed. A call to an accountant to sell them tax software is a B2B call even if you reached them on a personal mobile. A call to that same accountant to sell them a holiday package is a consumer call, on the same number. The classification hinges on why you are calling, not whose number is in the file.

Practically, staying inside the B2B lane still comes with obligations that predate August 2026 and do not go away: identify the purpose of the call up front, and honor an opt-out the moment someone asks. Fines for consumer violations have been reported up to €375,000, a figure that does not apply to properly scoped B2B calls, but only if the call actually is B2B by purpose, not just by which list it came from.

Step 1: audit where AI touches a prospect

List every point where something other than a human is generating a message a prospect sees or hears: AI-drafted cold email sequences, an AI SDR's voice calls, an auto-reply bot in LinkedIn DMs, a chat widget on a landing page. For each one, mark whether the AI's role is obvious to the person on the other end, or whether it is built to pass for a person. Only the second category needs a disclosure fix before August 2.

Step 2: write the disclosure line once

You do not need a different disclaimer for every channel. Write one short, honest line, something like stating up front that the call or message is AI-assisted, and drop it into every script and template that falls into the "not obvious" bucket from step one. Keep it short. The rule is about clarity at the start of the interaction, not a wall of legal text.

Step 3: check your list for consumer numbers

Pull your France-targeting list and check for personal mobile numbers mixed into what you are treating as a B2B file, and confirm your outreach purpose actually matches a business use case for every contact on it. Remember the accountant example above: the number does not decide the classification, the purpose of your call does. If a chunk of the list was scraped for a use case that is not clearly B2B, fix that before August 11, not after a complaint.

Step 4: ask your vendor who owns this

If you run an AI SDR platform or a dialer with AI voice on top, ask the vendor directly whether their system already inserts the required disclosure, or whether that is left to you to configure. Get the answer in writing before August 2. This is the one step people skip because they assume the vendor has already handled it, and it is the step most likely to bite if that assumption is wrong.

Before and after, side by side

Here is the practical shift for a B2B outbound team running calls and AI-assisted messaging into the EU and France specifically.

Where this fits into my own stack

I run Agent Frank, the AI SDR inside the Forge ecosystem, alongside Salesforge for sequencing and Leadsforge for the data feeding it, and I say that here as a disclosure, not a neutral recommendation. Ahead of both dates, I am doing exactly the four steps above on my own client campaigns: checking every voice and auto-reply touchpoint for whether the AI's role is obvious, writing one disclosure line and reusing it everywhere it is not, and re-checking France-targeting lists for anything that reads as consumer rather than B2B by purpose. None of this changes the 2000+ meetings booked for B2B clients number I can point to, it just keeps the campaigns that produced it inside the rules that apply to them now.

My take

I think the real risk this August is not the rules themselves, it is teams treating "B2B is exempt" as a blanket pass and not checking whether every call on a list actually is B2B by purpose, or assuming a vendor already built in AI disclosure when nobody ever confirmed it. Both fixes here take an afternoon, not a compliance project. Do the afternoon of work in July, and August is a non-event instead of a scramble.

FAQ