MK2 between the Palm and the Prompt

MK2 integrates generative tools into its post-production pipeline. What technology changes for the editing of a feature film selected at Cannes.

Six Cannes award-winning films in 2025, including the Palme d’Or. And alongside them, an AI lab, a dedicated festival, and artist residencies. MK2 is moving along two ridge lines that seem irreconcilable. Yet Elisha Karmitz, at the helm of the group founded by his father, refuses to take a side. Neither a starry-eyed technophile nor a temple guardian digging his heels in on 35mm. An embraced ridge line, between heritage and experimentation.

Your father, Marin Karmitz, built MK2 by often positioning himself "against the grain." In the face of AI, is your approach one of support, vigilance, or do you already see it as a defining feature of the cinema to come?

When my father talked about "standing against," it was not about being against something on principle, but about defending an idea and a worldview. On AI, our position is clear: we are neither for nor against it. AI is a tool, and more broadly, an industrial revolution. The real question, especially in creative industries, is not about the tool itself, but about the rules we set for using it and the ethics we choose to associate with it. The history of technology has already shown us this. Photography was used for propaganda and disinformation without calling into question its legitimacy as a medium. With AI, the phenomenon is similar, but amplified: extremely powerful creative tools are now accessible to the vast majority, which acutely raises the question of the veracity of information and fake news. At this stage, the problem is not AI as such, but the intention of those who use it. A portion of the investors and thinkers behind these technologies clearly view AI as a tool for control and the concentration of power. We have already experienced this scenario with the Internet, which carried a promise of freedom, but whose value was ultimately captured by a very limited number of players, at least in the Western world. It is therefore legitimate to ask ourselves collectively whether we want to replicate this model with AI, even as it represents a major civilizational issue. What is striking is that these questions remain far too marginal in the political debate. We should undoubtedly spend more time informing citizens, creators, and entrepreneurs, and clarifying public positions, particularly in France. The question of sovereignty is central, even if it remains difficult to conceive in a context where infrastructure—servers and networks—is overwhelmingly American.

Precisely on these issues of sovereignty and public policy: do you, at MK2, have discussions with institutions, such as the Ministry of Culture or the CNC?

Yes, these reflections do exist, particularly at the CNC, where highly competent teams are working on these issues. At MK2, we approach these questions from our field of expertise—creative industries, cinema, and the audiovisual sector—but it is clear that these issues go far beyond our sector and concern society as a whole.

Historically, MK2 has been associated with major auteurs, filmmakers like the Dardenne brothers, Abbas Kiarostami, or Pawel Pawlikowski. How do these auteurs perceive AI today, as well as MK2's involvement in this field?

Reactions are diverse, much like in society in general. Some filmmakers look at AI with interest and curiosity, others with concern or rejection. Many great directors have always paid close attention to technical developments. Jia Zhangke, for example, took an early interest in new forms of image-making and integrated AI into a festival he organizes in China. It is also worth remembering that technological disruptions have sometimes enabled major artistic breakthroughs. Abbas Kiarostami was able to create some of his most important works thanks to the arrival of the digital camera, which notably allowed him to bypass censorship in Iran. Filmmakers are aware of this: tools evolve, but the vision remains.

Yet, we see very sharp positions. Guillermo del Toro, for example, has publicly opposed AI. Isn't there a risk of a lasting fragmentation of the sector between supporters, opponents, and new entrants who use AI outside traditional channels?

Maybe. Are these positions reconcilable? I don't have a definitive answer. Today, technology is capable of reproducing forms with a quality almost equivalent to that of the human touch.

Let's take an example outside of cinema: a machine could make a Hermès bag with a quality almost identical to that of an artisan. But the value of that bag doesn't lie solely in its technical quality. It lies in the fact that it was made by hand, in the intention, the craftsmanship, the story that accompanies it. The question is similar for cinema. Maybe tomorrow, what will give a film its value won't just be its mode of production, but the way it was conceived and made. I'm not a futurist. I'm wary of those who claim to know precisely what the cinema of tomorrow will look like. In reality, nobody knows.

Given the uncertainty still surrounding the use of AI in many cultural organizations, how do you practically work at MK2? Are you testing these tools internally, or is this a matter delegated to the teams?

At MK2, we have set up a dedicated lab. It is a permanent workspace where we test tools, host events, and develop projects. Among other things, the lab teams created the Artefact AI Film Festival. Based on these experiments, we identify talent, whom we sometimes support through residencies or incubation. We then continue developing projects with them. At the same time, we produce a lot of internal educational content, particularly through a newsletter, and we are also developing our own tools. Today, workflows are starting to be well integrated at MK2.

On a personal level, do you use these generative AI tools?

No, not on a daily basis. I do not produce the content myself. In fact, this shows that having a good tool is not enough: above all, you need to have something to say. At the company level, we use these tools when it makes sense, within a clearly defined ethical