Part of the AI Regulation News hub.
France's newspaper alliance has asked the competition regulator to enforce commitments Google gave in 2022
The publishers are asking a regulator to apply undertakings Google gave in 2022, which they say run until 2027, to a feature that did not exist when those undertakings were given.
Bottom line: A complaint, not a decision. APIG referred the matter on 11 August 2026. The Autorite de la concurrence has not ruled, ordered anything, or found any breach on this record.
Who this affects: Publisher-side and platform-side counsel handling content licensing, competition lawyers watching how old commitments are stretched over new AI features, and any business whose content is ingested into an AI answer surface.
Issue date: Filed 11 August 2026. No deadline and no scheduled decision date on this record.
What changed: France's press alliance opened a competition front against generative AI answers, on the footing of neighbouring rights and existing binding commitments rather than any new AI statute.
Analysis: APIG's position is that France already has the tools, in the 2019 law and in commitments it says the regulator made binding. That route is faster than waiting for new legislation, which is why the filing is worth following whatever the Autorite decides.
Primary sources: APIG press release, 11 August 2026
- Instrument (EN)
- Saisine (referral) of the Autorite de la concurrence by APIG
- Authority
- Alliance de la presse d'information generale (complainant); Autorite de la concurrence (regulator seised)
- Jurisdiction
- France
- Status
- Filed. No decision, no published investigation and no order on this record.
- Bindingness
- None. A referral does not bind Google or anyone else. The 2022 commitments it seeks to enforce were, on APIG's account, made binding by the Autorite.
- Issue date / next deadline
- Filed 11 August 2026. No scheduled decision date.
- Primary source
- https://www.alliancepresse.fr/actualite/services-dia-de-google-lalliance-saisit-lautorite-de-la-concurrence/
What was filed, and what it asks for
The Alliance de la presse d'information generale, which represents French general-information publishers, says it has referred Google to the Autorite de la concurrence to make it respect the commitments it gave in 2022. The trigger is the deployment in France of AI Overviews and AI Mode.
Since that deployment, the alliance says, an AI-generated answer appears on the main search engine ahead of the classic links, and it is built in part from press publishers' content.
The complaint is about process rather than the technology. In the alliance's account the launch came before any transparent, good-faith negotiation was opened. Publishers were offered what it calls a simple update to the existing licence agreement, with the only alternative being a technical withdrawal that costs them visibility. The phrase it uses is that they were presented with a fait accompli.
APIG is explicit that this is not meant to end talks: the referral "ne ferme pas la voie a la negociation : elle vise a en retablir les conditions", which is to say it aims to restore the conditions for negotiation rather than close it off.
Why APIG says 2022 commitments reach a 2026 feature
The legal spine, as APIG sets it out, is French neighbouring rights. The alliance points to the law of 24 July 2019, which it says made France the first EU country to transpose the neighbouring-rights directive and which subjects use of press content to prior authorisation and to remuneration.
On top of that sits a competition layer. According to the alliance's account, the Autorite sanctioned Google in April 2020 for unfair practices and required it to negotiate in good faith with publishers, later fined it 500 million euros for failing to comply, and then made binding a set of Google commitments on good-faith negotiation, transparency and non-discrimination that run until 2027.
That is the point APIG asks the regulator to apply here. Its argument is that deploying a new use of press content without prior authorisation or dedicated remuneration disregards undertakings it says Google already gave.
The numbers and the neighbouring case, both as the alliance reports them
On scale, the release cites Arcom, the French audiovisual and digital communication regulator, as estimating the traffic loss attributable to AI-generated summaries at between 33 and 38 percent on European markets where these services are already active.
On direction of travel, the release points to a decision a month earlier. It says that on 8 July 2026, following an APIG referral lodged in September 2025, the Autorite issued four interim measures against Meta, and that at the press conference on that decision its president Benoit Coeure said protected content that would be used by AI Overviews must be remunerated under neighbouring rights. The conditional is his, and we have not opened that decision.
Both of those come to us through the complainant. We have not opened the Arcom material or the Meta decision, and we flag that rather than presenting either as verified.
What the alliance's president said
Marc Feuillee, the alliance's president, framed it this way: "Que l'intelligence artificielle se construise sur les contenus de la presse dit une chose simple : cette information a une valeur considerable. Les editeurs ne demandent pas d'arreter l'innovation. Ils demandent que cette valeur soit partagee et que l'utilisation de leurs productions soit remuneree, comme la loi l'exige."
In our translation: that artificial intelligence is built on press content says something simple, which is that this information has considerable value. Publishers are not asking for innovation to stop. They are asking for that value to be shared and for the use of their work to be remunerated, as the law requires.
The practical consequence
For anyone advising a publisher, the point that transfers is the legal theory. Where a platform already sits under negotiated or imposed undertakings about how it deals with content owners, a new AI surface may fall inside those undertakings without anybody legislating again.
For anyone advising a platform, the same question runs the other way. Commitments given in a search or licensing context, drafted before generative answers existed, may be read to cover a product built later. The alliance's argument is precisely that the obligation attaches to the use of the content, not to the interface that displays it.
Nothing in this referral binds Google. Its significance is that it tests whether existing neighbouring-rights rules and binding commitments can be applied to a later AI feature.
What we did not verify
We opened the alliance's own release and nothing else. It is a primary source for the fact and content of its filing, and it is the complainant's account for everything else in this piece.
One date we deliberately do not state. The release puts the 500 million euro fine in June 2022, while other coverage puts it in July 2021. We have not opened the Autorite's decision, so we report the fine without asserting when it was imposed.
We also make no claim about whether the Autorite will take the case up, how quickly, or what it might decide.
APIG is trying to reach a 2026 AI feature through neighbouring rights and undertakings it says the regulator made binding until 2027, a faster route than new AI legislation and one that transfers to any platform already operating under content-related commitments.
Source File
Open the APIG release of 11 August 2026 and confirm the referral, the reference to the 2022 commitments running to 2027, the Arcom range, the 8 July 2026 Meta measures and the Feuillee quotation.
Les editeurs ne demandent pas d'arreter l'innovation. Ils demandent que cette valeur soit partagee et que l'utilisation de leurs productions soit remuneree, comme la loi l'exige. ยท Marc Feuillee, president de l'Alliance de la presse d'information generale, 11 August 2026
FAQ
Has the regulator decided anything?
No. This is a referral to the Autorite de la concurrence, filed on 11 August 2026. On this record the Autorite has not ruled, has not published an investigation decision, and has ordered nothing. Nothing in the filing binds Google.
What exactly is being alleged?
That deploying AI Overviews and AI Mode in France, using press content without prior authorisation or dedicated remuneration, disregards commitments Google gave on good-faith negotiation, transparency and non-discrimination, which it says the Autorite made binding and which it says run until 2027. That is the alliance's allegation, not a finding.
Why would 2022 commitments reach an AI feature launched later?
Because the alliance's argument attaches the obligation to the use of press content rather than to the product displaying it. On that reading a new surface built on the same content falls inside undertakings drafted before it existed. Whether the regulator accepts that is exactly what is undecided.
What does this mean outside France?
Directly, nothing. It turns on French neighbouring-rights law and commitments specific to the Autorite. Indirectly it is a template worth knowing: where a platform already operates under content-related undertakings, a complainant may not need new AI legislation to raise the question.
Related briefings
Sponsored Training
Practical AI training for regulated professionals, built around verification, documentation and a defensible process. See the courses.