Cannes is back to being the gatekeeper of creatives

Last June a Brazilian agency won a Grand Prix at Cannes for Whirlpool's Consul brand. The case film had faked CNN Brasil news coverage. Cannes took back 12 awards.

This June, on a stage at the same festival, will.i.am asked for a new file for humans.

Between those two Junes the record business wrote down what counts as human work, put a mark on it, and made it a condition of entering a chart. At Nestle Pakistan more than 20,000 of our creatives went through a quality check before they aired, and not one of those checks asked who made the work. Nobody in advertising asks now either.

Cannes had to start checking whether the work was real

The agency was DM9, and it withdrew three entries carrying 12 awards between them, a Creative Data Grand Prix among them. Cannes said why in its own statement: "AI-generated and manipulated content was used within the case film to simulate real world events and campaign outcomes, resulting in the jury being presented with inaccurate information." The co-president resigned.

Read that quote again. The fraud was in the evidence about the ad.

Cannes wrote new rules within a fortnight. Companies "found to have wilfully submitted false or misleading work" can be banned for up to three years, sources are requested at entry, and an integrity handbook followed in November. Then entries fell to 20,050 from 26,900, down 25.46%. Anyone who says awards no longer matter should explain that number. Entry fees run EUR 690 to EUR 2,825 each, and agencies cutting staff every quarter still pay it, because a Lion is what they charge a client for next year.

will.i.am wants a file that proves you are human

On stage at Cannes, will.i.am said he wanted a "new file for humans", one carrying the meaning and intent of a piece and not only the audio. His reason was money. "If we're sharing the same file and it's valued at the same fraction of a penny, that's kind of inhumane because we're not valuing the human effort to make it." And: "A diamond is not valued the same as a freaking pearl."

He is right about the penny. A written song and a prompted one should not clear at the same rate.

Now the boring question. Somebody has to run that register, check the files and charge for them. Who. At what price. And what happens to the kid in Lahore with a story, a laptop and no file. Everybody calling for one came up through a system that already decided who got heard, and they did well out of it. He also said he fears "human greed" more than AI, which is the truest thing anybody said at that festival, and it points back at the stage he said it from.

Music at least had a theft to point at

Deezer published the numbers on 21 July. Nearly 90,000 fully AI-generated tracks a day in June, more than half of everything uploaded, taking 1% to 3% of streams. Then the line that explains every rule since: "up to 85% of the streams generated by fully AI-generated tracks were in fact fraudulent in 2025." That money came out of the royalty pool paying the artists Deezer already had.

So the marks make sense where they landed. Eight bodies agreed them eleven days before Deezer published, IFPI put "substantially human made" on chart eligibility nine days after, and the Grammys now rule on whether a human contribution was "meaningful and more than de minimis". A theft got a rule, and the rule handed four institutions the power to define the word "substantially" for everybody else.

Advertising's only line is one sentence in a European law

From August, the EU AI Act's Article 50 requires disclosure of synthetic content that looks like real people or events, with fines up to EUR 15mn or 3% of turnover, and exempts anything made under "substantive human review or editorial control". Lawyers drew that line where our industry should have. An ad shot with real people and then cut, graded, dubbed and versioned with AI sits outside it. An ad with a generated human inside it does not.

Three campaigns used synthetic humans and the only difference is who said yes. Cadbury cloned Shah Rukh Khan's face and voice with his permission, localised the ad to 2,500 small retailers and won a Clio. H&M built digital twins of 30 models who get paid for each use. The synthetic actress Tilly Norwood got the opposite reception, with SAG-AFTRA calling synthetic performers a replacement of human ones. Khan said yes and was paid. H&M's models said yes and get paid per use. Nobody said yes to Tilly Norwood, because there is nobody there to say it.

The shoot got cheap. The judgement did not

PJ Accetturo made Kalshi's NBA Finals spot alone in two days for about USD 2,000, and in his words "it took 300 to 400 generations to get 15 usable clips". The production companies Kalshi asked had quoted a "six or seven figure range". Four hundred generations, fifteen used: the job moved to knowing which fifteen are any good, which is what a creative director has always been for.

Meanwhile the people writing the rules are firing the judges. Omnicom's CFO told a Goldman Sachs conference the merged group goes from 128,200 staff to about 105,000 by the end of this year, with USD 1bn a year of its promised savings coming out of labour, some of it offshored to Colombia, Costa Rica and India. FCB folds into BBDO, DDB and MullenLowe into TBWA, leaving three networks where there were six. Cannes retired its Creative Company of the Year award in May because mergers meant it "no longer provides a consistent or comparable benchmark".

One agency cheated and now everybody shows papers

Every one of these bodies caught something real. Cannes caught a faked case film. Deezer caught a royalty skim. SAG-AFTRA is defending people whose faces get taken. All fair.

Look at what each one built on top of it. A handbook. Source material on demand. Entry at up to EUR 2,825. A chart place conditional on a phrase the chart's owner defines. A EUR 15mn fine that a holding company's legal department absorbs and a two-person studio in Karachi does not. One agency faked a case film, and the answer on offer is paperwork for everybody who was never invited in the first place.

None of it puts the tools back. Nobody un-built the factories, nobody policed the steam engine out of England, and no handbook takes video generation off a laptop in Lahore or Lagos.

So here is what is actually happening. Cannes and IFPI are trying to become the gatekeepers of creative work again, and they are doing it in the name of defending humanity. The real job on the table is deciding who gets called a creative, which is the last thing either of them still owns. Omnicom cannot make an ad cheaper than one person with a video model and a week, and it knows it.

The only verdict that ever counted is whether real people want to watch the work. That verdict now comes back in a day, in public, from the audience, and Cannes cannot issue it, sell it or charge for it. So Cannes is building something it can charge for instead.

The fraud is the pretending. DM9 pretended a case film was news. Tilly Norwood has nobody there to agree to anything. Catch that, all day, and hand out the three-year bans while you are at it. That is a different job from asking a storyteller to prove she is human.

Fruit Love Island did 300mn views without entering anything

In March a show called Fruit Love Island appeared from an account nobody had heard of, passed 3mn followers in nine days and finished on more than 300mn views. No studio, no commission, no pretence about how it was made. Oksana Bondarieva publishes her own numbers for an AI-native animated series she makes alone: 4,411,496 views in 90 days, 97.2% of them from people who do not follow her, an episode written and cut in about thirty minutes. The animated feature Critterz was made for under USD 30mn in nine months against the several years and USD 100mn-plus a studio takes.

What kept people out of Pixar was four years and USD 100mn. What keeps them out now is a laptop and a story, which keeps nobody out.

Some of it lands badly, and that pushback is honest. An AI film called "Thanksgiving Day" won cinema screenings and was pulled after the backlash, and Tropfest's own alumni told the festival its AI finalist devalued their work. That is an argument about taste and about jobs, and it deserves having. It is a different argument from fraud.

YouTube asks whether the work is a con. Cannes asks whether you are human.

In January YouTube terminated 16 channels holding 35mn subscribers and 4.7bn lifetime views under its rules on spam and deceptive practices. Those rules say nothing about AI and nobody was asked to prove they were human. The question was whether the work was a con.

That is the whole test, and it leaves a paper trail. Who made this, and did the people in it agree to be in it. Khan agreed and got paid. H&M's models agreed and get paid per use. Nobody agreed to Tilly Norwood.

The other question, whether a machine touched the work, is one nobody can answer and soon nobody will bother to ask. "AI-Assisted" already covers anything "created substantially by humans", which is most creative work now and almost all of ours. Certify that, charge for the certificate, and you have sold admission to the craft.

Cannes lost 6,850 entries in a year and retired its own company of the year award because it could no longer tell who was who. In the same twelve months 300mn people watched a show made by an account nobody had heard of, 4.4mn watched a woman working alone in her own room, and a feature film got made for less than a third of what a studio spends. Nobody asked any of those audiences for a file, and nobody had to, because they were the test.