Hassabis Tells Cannes Lions True Creativity Is the AGI Bottleneck

Demis Hassabis told a room of brand and agency leaders at Cannes Lions on Wednesday that the missing pieces keeping AI short of human-level intelligence are not bigger models but something older and harder to engineer: true creativity. The Google DeepMind co-founder and CEO, on the Lumière Theatre stage in conversation with Bloomberg TV’s Francine Lacqua, framed the next 10, 15 years as the period in which generative tools become collaborators an artist can steer in plain language.

The chat, titled The Future of Creativity with Demis Hassabis and listed on the Cannes Lions session page, sat two days after Hassabis’s own lab confirmed a $75 million equity investment and research partnership with A24, the independent film studio.

Hassabis Tells Cannes True Creativity Is What AGI Still Lacks

Hassabis ran through DeepMind’s standard definition of artificial general intelligence, then said the gap was narrowing but not closed. He named two missing pieces: long-term reasoning and planning, and creativity that goes past remixing. The framing came in a fireside chat moderated by Lacqua, with Scott Belsky, founder of A24 Labs and Adobe’s former chief strategy officer, joining later in the discussion.

On creativity, Hassabis drew a line between what today’s generative tools already do and what he thinks AGI will require. Iterating on a known idea or remixing existing things, he suggested, is work current systems already do well. Coming up with something genuinely novel is the work current systems still struggle with. He tied that gap directly to the AGI roadmap.

If you didn’t like some part of the image or you wanted to adjust it, you could just describe in natural language explaining what you want.

That back-and-forth, the loop between a creator and a model that responds to plain-language edits, is the part Hassabis said will define the next phase. Multimodal tools that generate and edit text, images, audio and video from a single conversation are where he sees real creative room, not in any single model release.

The $75 Million Handshake That Was Not on the Programme

Two days before the chat, on 22 June 2026, Google confirmed its first equity stake in a film studio: $75 million into A24, paired with a research partnership that gives the studio and its filmmakers a working channel into DeepMind’s models. The $75 million A24 partnership was confirmed as Cannes opened, per Tech Times. A24 gets access to DeepMind’s research infrastructure, and DeepMind researchers sit alongside A24 productions.

A24 is the kind of studio the panel was framed around. The partnership gives its filmmakers a working channel into DeepMind’s models and gives DeepMind researchers a seat inside A24 productions.

Adobe, the other company represented on stage through Belsky, runs the LIONS Creators programme in partnership with Cannes Lions for the third year running. The festival’s own programme copy puts Hassabis’s chat, Adobe’s creator track and the A24 deal inside the same five-day window.

The pattern echoes earlier platform shifts: the labs pick the studios they want feedback from, the studios get privileged access, and the rest of the industry reads about it after the fact. Whether that is a good deal for independent creators outside the A24s and Adobes is a question the panel did not address. Belsky’s argument on stage, that technological shifts have historically created new categories of jobs and creative opportunities, does not directly answer it either. WPP’s Cannes Lions trust push, announced earlier in the week, was a parallel attempt to define the ground rules for AI in agency work.

Why Scott Belsky Wasn’t There as a Commentator

Belsky joined the chat from a position the festival bio understated. He founded Behance, ran Adobe’s product and design organisation for years, and now runs A24 Labs, the independent film studio’s in-house incubator. He joined Cornell Tech’s council in September 2025, per the university’s announcement, in a role that gave him a standing seat on questions about AI and creative work.

He is also the executive most directly inside the company that just signed the $75 million partnership with DeepMind. A24 Labs is the unit inside A24 that handles exactly the kind of AI-tooling work the panel was describing. When he told the room that AI might give rise to entirely new forms of entertainment and storytelling, he was not speaking as a sideline commentator. He was speaking as the operator who would help decide which new forms his company builds first.

What ‘True Creativity’ Means in Hassabis’s AGI Math

Hassabis’s creativity argument is not marketing language. He has been pushing it for at least four months as a load-bearing piece of his AGI framework. At the India AI Impact Summit in New Delhi in February 2026, he said the same gap, jagged intelligence and missing creativity, was the reason AGI is still five to eight years away, per the Observer’s write-up of his appearance there.

On Wednesday he widened the point. Creativity, in his telling, is the ability to come up with something novel, not to remix what already exists. That is a higher bar than producing competent work, and it is the bar Hassabis has set for his own lab. He tied that bar to a separate argument about the human brain’s efficiency and complexity, including emotions, creativity and dreaming, things current AI does not model well.

Google DeepMind’s working AGI definition has not changed. A system that exhibits all the cognitive capabilities humans are capable of remains the company’s long-stated target.

What the Cannes chat added was the part about how those capabilities would actually be used. The collaboration mode Hassabis described, an artist steering a model in plain language, is the surface where the missing creativity gap either closes or stays open. He has been making versions of this case for a while, including at the India AI Impact Summit in February. Cannes gave him the room to make it to a marketing audience instead of a summit one.

Overhyped Near Term, Underappreciated Over the Next 15 Years

Hassabis is also willing to undercut the AI hype cycle when the venue allows. On Wednesday he said AI is overhyped in the very near term and underappreciated over the next 10 to 15 years, a stance he has taken before but rarely from a festival main stage.

I think AI is perhaps a bit overhyped in the very near term. But then over the medium long term, over the next 10, 15 years, it is still underappreciated how much transformation AI is going to create.

The framing matters because the people in the room were not buying AI tools. They were planning ad campaigns, agency rosters and brand strategy for the next 18 months. Hassabis told them their planning horizon was wrong. The transformation, in his telling, will run longer than any campaign cycle and hit harder than current product roadmaps suggest.

He went further. The cumulative effect of the next decade and a half of AI progress could be, in his words, almost a new human era. That is a longer claim than the standard CEO forecast of a year or two out, and it positions DeepMind’s lab work as the foundation of that era.

The Watermark That Has to Outlive the Panel

The last topic Hassabis covered was the one that may matter most after the cameras leave. Asked about trust, he pointed to the watermarking technology DeepMind developed and bundled into its models from the start. The argument is that generated content should be traceable through common industry standards, not left to each platform to label however it likes.

• Cannes Lions 2026 runs 22 to 26 June with around 500 speakers and more than 150 hours of programming, per the festival’s programme release.
• Google’s $75 million A24 equity investment is the search giant’s first film-studio stake, per Tech Times (23 June 2026).
• DeepMind’s SynthID watermarking expanded to text in Gemini and video in Veo on 14 May 2024, per the DeepMind blog.
• Hassabis placed AGI five to eight years away at the India AI Impact Summit in New Delhi on 18 February 2026, per the Observer.

SynthID embeds imperceptible signals inside AI-generated text, images, audio and video at the point of generation. The SynthID text and video watermarking post from May 2024 describes how the watermark survives cropping, mild paraphrasing and pixel-level edits, but loses confidence under heavy rewriting or translation. Common industry standards would mean every major lab’s models carrying compatible signals.

That is a separate fight from the one Hassabis spent most of the chat on, but it is the one his lab is actually shipping. Global AI rules, the subject of Sam Altman’s separate appeal to G7 leaders covered in earlier Cannes governance reporting, will not land soon enough for the 2026 campaign cycle. Watermarking, if DeepMind’s standards get adopted, would.

Frequently Asked Questions

When did Hassabis speak at Cannes Lions and what was the session called?

Hassabis spoke at the Lumière Theatre on Wednesday, 24 June 2026, from 14:30 to 15:00 local time. The session, titled The Future of Creativity with Demis Hassabis, was part of the festival’s Innovation Unwrapped content stream and was moderated by Bloomberg TV anchor Francine Lacqua.

What did Hassabis mean by ‘true creativity’ in his AGI definition?

He defined it as the ability to come up with something genuinely novel, distinct from iterating on a known idea or remixing existing content. He named it, alongside long-term reasoning and planning, as one of two pieces still missing from what Google DeepMind counts as AGI.

What is the Google DeepMind-A24 partnership and when was it announced?

Google invested $75 million in A24, its first equity stake in a film studio, alongside a multi-project research partnership that gives A24 access to DeepMind’s research infrastructure. The deal was announced on 22 June 2026, two days before the Cannes fireside chat.

How does SynthID watermarking work and which Google products use it?

SynthID embeds imperceptible signals into AI-generated text, images, audio and video at the point of generation. DeepMind’s blog confirms it has been integrated into the Gemini app and web experience for text and into Veo for video. The signal survives mild edits but loses confidence under heavy rewriting or translation.

What did Belsky add to the conversation?

Belsky, founder of A24 Labs and former Adobe chief strategy officer, argued that AI will give rise to new forms of entertainment and personalised storytelling, and that historical technology shifts have created new categories of jobs. He did not address concerns from artists who remain skeptical of generative tools.