Film summit coverage on 7 September 2026 carried one quote around the world, and it was a good one. Emmanuel Macron, standing at the Fondation Maeght in Saint-Paul-de-Vence, told a room of more than 200 screen-industry figures that “we would be making a tremendous mistake if this technology became a tool that replaces, or claims to replace, creators.” South Korea’s president Lee Jae Myung, co-chairing, said the industry “must be rooted in human creativity and must resonate with human emotions.” Wire services ran it, aggregators reran it, and the story became a warning about artificial intelligence delivered by two heads of state.
The warning was real. It was also the least binding thing that happened that day. Underneath the speeches, the Lumière Summit produced six actual documents — one overarching declaration and five thematic ones — and those documents have signatures on them. The signatures are where the story is, because they are countable, and counting them tells you something the speeches do not.
The Saint-Paul-de-Vence Declaration on Artificial Intelligence, the text that would commit signatories to paying rightsholders when protected works train a model, was signed by four named parties: France, the Republic of Korea, and two French collecting societies. No studio signed it. No streamer signed it. No company that builds machine learning systems signed it, for the straightforward reason that none of them were in the building. Meanwhile the Motion Picture Association — the trade body for Netflix, Disney, Sony, Universal, Paramount, Warner Bros. Discovery and Amazon MGM — signed two of the six documents, and the AI one was not among them.
This article walks through what the film summit actually produced, who put their name to which text, and what the signature pattern implies for anyone buying, deploying or commissioning AI systems that touch other people’s copyrighted work.
Table of contents
- What Actually Happened at the Film Summit in Saint-Paul-de-Vence
- The Film Summit’s Real Output Was Six Documents, Not Two Speeches
- Counting the Signatures on the Film Summit’s AI Declaration
- The Film Summit’s AI Declaration, Principle by Principle
- What Hollywood Signed at the Film Summit, and What It Did Not
- The Money the Film Summit Announced, and How Far It Goes
- The Second Argument the Film Summit Made and Did Not Paper
- The Film Summit’s Other Four Declarations Are Not Filler
- What the Film Summit Means If You Deploy AI on Content
- What the Film Summit Left Unresolved
- Frequently Asked Questions About the Film Summit
- References
What Actually Happened at the Film Summit in Saint-Paul-de-Vence
The event was the first of its kind, and its framing was deliberately governmental rather than commercial.
Two presidents, one medieval village
The film summit convened on Monday 7 September 2026 at the Fondation Maeght, a contemporary art foundation in the medieval village of Saint-Paul-de-Vence, inland from Nice. It was co-chaired by Macron and Lee, and named for Auguste and Louis Lumière — a choice that signals the organisers wanted the framing to be about cinema as a civilisational form rather than about quarterly output.
Attendance ran past 200 industry figures drawn from 55 countries. George Clooney was there in person. Pedro Almodóvar and Juliette Binoche took part remotely during the launch. Executives from Sony Pictures and Disney attended, as did representatives from Netflix and TikTok.
Who was not in the room
The absences are as informative as the guest list. YouTube did not attend. Meta, and therefore Instagram, did not attend. And the major generative AI developers — the companies actually training the models that the film summit spent the day worrying about — were not present either.
That matters for how you read the day’s output. A summit can only bind the people who turn up and sign. When the parties whose behaviour a declaration is designed to change are not in the room, the declaration becomes a statement of the position that other parties will now negotiate from.
| Element | Detail |
|---|---|
| Date | Monday 7 September 2026 |
| Venue | Fondation Maeght, Saint-Paul-de-Vence, near Nice |
| Co-chairs | Emmanuel Macron; Lee Jae Myung |
| Attendance | 200+ figures from 55 countries |
| Documents adopted | One overarching declaration, five thematic texts |
| Headline funding | €1 billion over five years, 2027–2031 |
| Present | Netflix, Disney, Sony, TikTok, MPA leadership |
| Absent | YouTube, Meta, major generative AI developers |
The Film Summit's Real Output Was Six Documents, Not Two Speeches
Reporting on the day split cleanly. The wire copy covered the warning; the trade press covered the paperwork. The paperwork is the part with legal shape.
One chapeau text and five thematic declarations
The film summit adopted a general Lumière Declaration on the Future of Cinema and the Moving Image, plus five thematic declarations covering artificial intelligence, piracy, gender equality, image education and environmental sustainability. Countries, screen-industry bodies and rights organisations were the signing parties.
All of the thematic declarations remain open for additional signatures. That is a deliberate design: the texts are meant to accumulate names over time rather than to close on the day. It also means today’s signature count is a floor, not a verdict — but it is the only count that exists right now.
Why the declarations outrank the warning
A presidential warning has no addressee. A declaration has a signature block, a set of commitments, and a list of parties who can later be asked whether they are honouring it. If you want to know what the film summit will still mean in 2028, read the signature blocks rather than the transcripts.
| Document | Subject | Named signatories in reporting |
|---|---|---|
| Lumière Declaration on the Future of Cinema and the Moving Image | Overarching principles | Motion Picture Association |
| Saint-Paul-de-Vence Declaration on Artificial Intelligence | AI and rights | France; Republic of Korea; SACEM; Adami |
| Declaration on Strengthened International Cooperation in the Fight Against Piracy | Enforcement | France; Republic of Korea; Motion Picture Association |
| Declaration for Safer Sets and Greater Gender Equality | Workplace conduct | Not individually itemised in coverage |
| Manifesto for Universal Image Education | Image literacy | France; Italy |
| Declaration Towards More Sustainable Screen Industries | Environment | Not individually itemised in coverage |
Counting the Signatures on the Film Summit's AI Declaration
This is the arithmetic that no coverage of the day performed, and it takes about a minute.
Four names, and what each one is
The Saint-Paul-de-Vence Declaration on Artificial Intelligence carries four named signatories. Two are sovereign states: France and the Republic of Korea. The other two are French collective management organisations. SACEM administers rights for music authors, composers and publishers. Adami collects and distributes on behalf of performers — actors and musicians whose recorded performances get reused.
Both are serious institutions with real leverage over French licensing. Neither is a film studio, a streaming platform, or a company that trains models. So the text asking for payment when protected works train AI systems was signed by two governments and two rights collectors, and by nobody who would be paying.
The category that is missing entirely
Group the named signatories by what kind of organisation they are and the gap is stark. Governments: two. Collecting societies: two. Film and television companies: zero. Generative AI developers: zero.
Why this is not a gotcha
It would be easy to read the gap as hypocrisy, and that reading would be wrong. Governments sign declarations of intent because that is the instrument available to them before legislation. Collecting societies sign because their mandate is precisely to assert that reuse requires payment. Studios do not sign texts that create licensing obligations before they know the price. AI developers do not sign texts they were not invited to negotiate.
The pattern is what an early-stage position looks like, not what a failure looks like. But the practical consequence is unchanged: on the day the film summit warned about AI, the AI commitments were made by parties who do not build AI.
The Film Summit's AI Declaration, Principle by Principle
The text sets out six principles. Read as procurement requirements rather than as diplomacy, they are surprisingly concrete.
Payment, transparency and employment
The declaration places AI systems inside the scope of UNESCO’s 2005 Convention on the Protection and Promotion of the Diversity of Cultural Expressions, which is an existing binding treaty rather than a new instrument — a smart move, because it borrows enforcement architecture that already exists.
It then asks for fair payment to rightsholders when protected works are used to train AI systems, and for the preservation of human artistic and creative employment as “an enduring public policy objective.” That second phrase is doing heavy lifting: it frames creative jobs as a policy target a state may legitimately protect, which is the predicate for future subsidy or quota rules.
The transparency asks are the operational ones
Three transparency commitments sit in the text: revealing the corpora a model was trained on, identifying AI-generated content, and enabling rightsholders to track where their work has been used. Each of those is a technical capability with a cost, and each is already appearing in commercial contracts independently of any declaration.
That third commitment is the hardest of the set. Tracking reuse across a trained model is not a lookup; it is an inference problem closer to computer vision than to database retrieval, and no standard method for it exists. The film summit asked for the capability without specifying who builds it.
The final two principles affirm the sovereign authority of nations to design their own cultural policies, and support AI development in underrepresented languages while extending rights protections across the whole creative chain.
| Principle | The question it forces on a buyer |
|---|---|
| Covered by the 2005 UNESCO Convention | Does cultural-diversity law now reach our AI vendor? |
| Fair payment for training on protected works | Can our vendor evidence a licence for its training data? |
| Human creative employment as policy objective | Will public funding start carrying headcount conditions? |
| Disclosure of the training corpus | Will the vendor tell us what the model was trained on? |
| Identification of AI-generated content | Do our outputs carry durable provenance markers? |
| Rightsholder tracking of reuse | Can a third party audit our corpus for their work? |
What Hollywood Signed at the Film Summit, and What It Did Not
The Motion Picture Association’s behaviour at the film summit is the single most legible signal of where the industry actually stands.
Two of six, and the two it chose
The MPA signed the Lumière Declaration on the Future of Cinema and the Moving Image, and the Declaration on Strengthened International Cooperation in the Fight Against Film and Audiovisual Piracy. That is two documents out of six. It did not sign the artificial intelligence declaration.
Charles Rivkin, the MPA’s chair and chief executive, framed the day in terms of common ground: “From the U.S. to France to Korea and beyond, we each have a stake in ensuring that our industry remains a force for openness, a source of investment, and a driver of jobs, growth, and opportunity.” Rivkin and member studio executives joined roundtables with both presidents, and Emilie Anthonis, the MPA’s president for Europe, the Middle East and Africa, took part in the anti-piracy sessions.
The selection is economically coherent
Look at what the two signed texts do. The piracy declaration commits signatories to four lines of action: a permanent multilateral body linking regulatory and enforcement agencies, stronger intelligence-sharing on illicit services, deeper cooperation with online intermediaries on site-blocking, and baseline standards so an enforcement action in one country carries weight in another. Every one of those protects studio revenue at another party’s expense.
The AI declaration does the reverse. Paying for training data is a cost. Disclosing datasets is a disclosure risk. Rightsholder tracking is an audit obligation. The MPA’s members are not only buyers of AI tooling; several are also building it internally. Signing that text would have created obligations pointing directly at their own systems.
Seven companies, none of them bound
The MPA represents Netflix, Paramount Pictures, Prime Video and Amazon MGM Studios, Sony Pictures, Universal Studios, The Walt Disney Studios and Warner Bros. Discovery — seven of the largest content producers and distributors on earth. Through the piracy declaration, all seven are now associated with a cross-border enforcement commitment. Through the AI declaration, none of them are associated with anything at all.
The Money the Film Summit Announced, and How Far It Goes
Alongside the paperwork came the number that led most headlines.
One billion euros, split down the middle
Macron and Lee launched what France called the Lumière Partnership and Korea called the Global Film and Video Industry Investment Cooperation Initiative: €1 billion, roughly $1.16 billion, over five years. Each country puts in €500 million. France’s half comes from Bpifrance, the state investment bank. Deployment starts in 2027 and runs to 2031.
Lee described the goal as backing “local content that people can relate to across borders and language barriers,” with a responsibility for “mutual growth and coexistence between the local and the global.” Macron’s framing was blunter: “We must not fear any of these changes, but must not be swayed by them either. No country can build the future alone.”
The independent cinema body is a rounding error
Separately, Macron announced a new organisation to support independent cinema with a budget of €30 million. Set against the billion, that is 3%. Set against the annualised partnership spend — €1 billion over five years is €200 million a year — it is 15% of one year.
Neither figure is trivial in absolute terms. But the film summit’s own money is aimed at production and distribution capacity, not at the rights-and-training problem the AI declaration describes. There is no fund attached to the AI text.
The Second Argument the Film Summit Made and Did Not Paper
The most quoted line was about replacement. The more unusual argument was about distribution.
“Our attention is an economic resource”
Macron told the film summit that the competition for attention run by AI systems and social platforms marks a “genuine change of era” for how cultural works are made and distributed. Audiences, he argued, are now fed a constant stream of material engineered to hold them for a few seconds rather than to “surprise, move or make people think.”
His prescription used a phrase with obvious political history: taking back control. “Taking back control means reclaiming attention, putting the work before the flow, and protecting those who create.”
This is a different problem from training data
It is worth separating the two arguments, because they have different remedies. The training-data argument is about inputs, and it resolves into licensing, disclosure and payment. The attention argument is about outputs and ranking, and it resolves into platform regulation, recommendation transparency and distribution rules.
The film summit’s AI declaration addresses the first. It does not really address the second — and the platforms that would be the subject of any attention remedy, YouTube and Meta in particular, were not present to be asked.
The Film Summit's Other Four Declarations Are Not Filler
The AI text drew the coverage, but the remaining declarations contain the more immediately actionable commitments.
Piracy, gender, education and sustainability
The piracy declaration is the one with the most machinery, because cross-border enforcement is a solved problem in structure if not in practice. The gender text commits signatories to equal opportunity in employment, financing, festivals and pay; to stronger harassment prevention and reporting; and to greater representation of women and under-represented groups in creative and decision-making roles.
France’s Centre national du cinéma has conditioned public funding on harassment-prevention compliance since 2021, and the United Kingdom established the Creative Industries Independent Standards Authority — so the gender declaration is codifying an approach with live national precedents rather than inventing one.
The education manifesto has the boldest ask
The Manifesto for Universal Image Education, a joint France–Italy initiative, calls for image literacy to be treated as a foundational twenty-first-century competency, and urges UNESCO to develop a binding international instrument specifically covering it. A binding instrument is a far stronger ask than anything in the AI declaration.
The sustainability declaration aims at convergence between the sustainability tools already in use across different markets, with shared European standards to be developed through CEN, the European standards body. Convergence work is unglamorous and tends to matter more than the announcements that accompany it.
What the Film Summit Means If You Deploy AI on Content
Most organisations reading this are not studios. The film summit still moves three things that show up in ordinary technology procurement.
Training-data provenance becomes a contract question
The declaration’s payment and disclosure principles are the polite version of a question your legal team will eventually ask a vendor: what was this model trained on, and do you have the right to have trained it? That question is already appearing in enterprise agreements. Treat it as a standing item in vendor management rather than a one-off due-diligence exercise.
The realistic near-term answer from most vendors is partial. Ask anyway, record the answer, and note where the vendor declined — the record is what protects you if the licensing position changes later.
Output labelling stops being optional
Identifying AI-generated content is the cheapest of the six principles to implement and the one most likely to become a hard requirement first, because it is verifiable. If you generate marketing, video, audio or images at scale, decide now whether provenance metadata survives your pipeline. It usually does not survive re-encoding or a trip through a design tool.
Nothing here is in force yet
None of this is law. A declaration is not a regulation, and the AI declaration in particular has no fund, no enforcement body and no signatory that builds models. The honest planning assumption is that these principles describe the direction of travel for European and Korean policy over the next two to three years, not the compliance position for this quarter. Organisations reviewing how AI fits their digital strategy should file it as a trend line, not a deadline.
What the Film Summit Left Unresolved
Three open questions come out of the day, and all three are answerable by watching the signature blocks.
Will the studios sign the AI text later?
Every thematic declaration remains open for signature. The MPA’s decision to sign two of six on day one is a starting position, not a final one. If a studio or a streamer adds its name over the coming months, that is a genuine shift and worth noticing. If the AI declaration’s signature list is still governments and collecting societies in a year, that tells you the licensing price was never agreed.
Can a declaration reach parties who were absent?
The generative AI developers were not at the film summit, which means the text describing their obligations was written entirely without them. Declarations of this kind usually work by becoming the reference point for later legislation, and France and Korea both have the legislative capacity to convert principles into rules. The route runs through national law, not through the declaration itself.
Does the money go where the warning pointed?
The €1 billion is for production and distribution capacity. The warning was about creative labour being displaced by systems trained on existing work. Those are related but not identical problems, and no announced funding attaches to the second one. Watching whether the 2027 deployment carries any AI-related conditions will be the clearest signal of whether the film summit’s rhetoric and its chequebook end up pointing the same way.
Frequently Asked Questions About the Film Summit
What was the Lumière Summit?
A first-of-its-kind international film summit held on 7 September 2026 at the Fondation Maeght in Saint-Paul-de-Vence, France, co-chaired by Emmanuel Macron and South Korean president Lee Jae Myung, with more than 200 screen-industry figures from 55 countries attending.
What exactly did Macron say about AI?
That “we would be making a tremendous mistake if this technology became a tool that replaces, or claims to replace, creators.” He separately argued that attention is “an economic resource” and called for “taking back control” from AI systems and social platforms.
Who signed the AI declaration?
France, the Republic of Korea, and the French collecting societies SACEM and Adami. No film studio, streaming platform or AI developer was among the named signatories.
Did the Motion Picture Association sign it?
No. The MPA signed the overarching Lumière Declaration on the Future of Cinema and the Moving Image, and the anti-piracy declaration. It did not sign the artificial intelligence text.
How much money was committed?
€1 billion, about $1.16 billion, over five years from 2027 to 2031 — €500 million from each country, with France’s share coming from Bpifrance. Macron separately announced a €30 million body for independent cinema.
Is any of this legally binding?
Not directly. These are declarations and a manifesto, not treaties or regulations. Their practical force comes from being the reference text that later national legislation in France, Korea and the EU can build on.
References
- Screen Industries Adopt Five Thematic Declarations at Lumiere Summit — Variety’s breakdown of all five thematic texts and their signatories.
- Macron, South Korea leader warn at film summit of AI risks to creativity — the AFP wire report, with the “tremendous mistake” quote and the attendance detail.
- MPA Signs Two Lumiere Summit Declarations — which documents the Motion Picture Association put its name to, and Charles Rivkin’s statement.
- Emmanuel Macron Calls for Global Initiatives to Protect Film — the attention-economy remarks and the Lumière Partnership announcement.
- 2005 Convention on the Protection and Promotion of the Diversity of Cultural Expressions — the UNESCO instrument the AI declaration places AI systems inside.
- Korea, France Each Pledge 500 Million Euros for Film Industry — the Korean-side account of the funding split and Lee’s remarks.
More AI coverage: explore Progressive Robot's AI Models, Tools & Releases hub — hands-on reviews, setup guides and benchmarks in one place.