During Goa’s Waves Film Bazaar, head of Cannes Next Sten-Kristian Saluveer framed 2025 as a hinge-year for screen industries grappling with artificial intelligence. Addressing a packed Knowledge Series session, he argued that the velocity of investment, policy and user adoption has pushed AI beyond the status of a passing trend.

Governments across the US, Europe and India, he noted, have shifted to strategic positions, and private capital has poured into model training, infrastructure and tooling. “It’s much more than a fad,” he explained, urging the room to “start thinking about it as a platform” comparable to earlier technological initiatives which enabled non-linear editing, mobile, social and streaming.

(El artículo continúa más abajo - Inf. publicitaria)

He underscored adoption with familiar tech-history markers. Whereas Netflix took years to reach 100 million subscribers, ChatGPT achieved a similar figure in five days, and daily usage at scale now shapes audience behaviour and expectations. In Saluveer’s opinion, we’re only a few years into the current generative wave: “three years,” he reminded us, but the platform is already maturing and altering how content can be produced, co-produced and distributed.

To help filmmakers frame their own responses, he offered a historical analogy: the scepticism which greeted Avid and, later, Final Cut Pro in the 1990s. Resistance eventually gave way to ubiquity. AI, he suggested, will follow a similar curve, especially since machine-learning features are now embedded within the tools filmmakers already use, from DaVinci to Premiere and After Effects. The practical lesson was to stop treating AI as an alien bolt-on and, instead, treat it “as software”: something to be configured, compared and tuned for specific tasks.

Saluveer also counselled nuance around creative choice. Citing the way directors choose between 35mm, 65/70mm, Alexa or RED, he posited that AI will become another creative option - powerful in some contexts, unsuitable in others. With automated workflows scaling content at unprecedented rates - he cited claims that 52% of written online material is already shaped in some way by AI - the decisive question becomes: how does a film cut through the noise? Hollywood pragmatists might claim to solve discovery with more algorithms, he said with a smile, but visibility strategies, marketing craft and distribution partnerships will matter more than ever.

Regulatory and industrial guardrails are, in his view, starting to take shape. He highlighted SAG-AFTRA’s evolving framework on synthetic replicas and consent, pointing to Echo Hunter - a generative-heavy production he claimed was the first to receive SAG clearance - as a sign that compliant, performer-respecting workflows are now possible. The core takeaway was succinct: “Yes, you can use synthetic replicas provided that you have consent and compensation.” He encouraged producers to read guild guidelines vis-à-vis performers, writers and directors, noting that the sector’s stance is increasingly “yes, but…” rather than “no”.

He also flagged Netflix’s recently published GenAI guidelines as a watershed: a major platform has now set out what is acceptable for mood boards, storyboards, background elements, VFX and pre-visualisation, and what requires heightened clearance when human performance or third-party IP is implicated. He predicted these checklists will travel fast through broadcaster and streamer ecosystems and will soon sit alongside release forms and delivery specs. Errors and omissions (E&O) coverage, copyright chain-of-title and audit trails, he added, will become non-negotiable for international distribution.

Festival gatekeepers, he said, are shifting as well. If last year saw caution, this year brought a cadre of titles - he name-checked Rotterdam, Toronto, San Sebastián and IDFA - which use generative tools to varying degrees.

To demystify the practice, Saluveer proposed three working “buckets”. First, he mentioned AI-assisted tools: text-based tools which accelerate development, research and administrative tasks across pre-production, production and distribution. He then spoke about AI-enhanced tools: post-production augmentation, especially for VFX and clean-ups. And, lastly, he touched upon AI-native tools, namely end-to-end workflows designed around generative media.

He then demonstrated “assistant” prompts built for real-world bottlenecks. One of these mapped public funds and minority co-production schemes for an Indian-European project trying to close a €250,000 gap, outputting eligibility criteria, requirements and likely timelines in minutes rather than days. A second reviewed line budgets to flag inconsistencies and compliance issues when multi-country co-producers submit different formats. A third, geared towards festival strategy, scraped deadline calendars and cost structures.

All of this, he stressed, belongs within a professional governance frame. Companies should decide which tools are approved, document data policies, and avoid “shadow AI” - staff uploading sensitive materials to unvetted services. Above all, teams should invest in prompting literacy: “the quality of the answer is determined by the quality of the question.” He pointed participants to free fluency courses and summarised the essentials of a good brief: context, role, action, format and tone.

Shifting to visuals, he revealed that production workflows are changing as image-led ideation becomes commonplace. He referenced a Barcelona-based firm, Artefacto Films, whose 14-step pipeline shows teams reverse-engineering from look-books and concept frames back to script iterations. He also screened a recent advertisement, for Liquid Death, largely created with generative tools, using it to temper panic with pragmatism: “This is the worst it will ever get,” he quipped, suggesting that commercials, shorts and indie VFX are the front lines where quality is already serviceable and rising.

The final pivot concerned legality and provenance. Courts in Europe and the US, he explained, are still emphasising that purely automated outputs are not copyrightable; human authorship must be evidenced. He advised producers to maintain detailed logs of AI use across video, audio and images, including the models employed, source assets and - crucially - the exact instructions. Without such documentation, films risk refusal by broadcasters, sales agents, platforms and festivals.

Saluveer closed with three “commandments.” First, adopt a strategic view and decide where AI belongs in your process rather than chasing every new tool. Second, be realistic about capabilities: AI can feel like “1,000 production assistants,” but you can’t finish your feature without craft and judgement. Third, invest in people and policy - consent, compensation, copyright, security - so that creative experimentation sits on firm legal ground.