Whoever publishes AI content now owns what they cannot vouch for, and the law is pricing that exposure.
Synthetic output has become the raw material of creative and media. A third of video ad assets were made with generative AI in 2026, and the share is still rising. Most of it goes out with nothing to show where it came from.
From August 2026 that stopped being a private choice. California and the European Union both began requiring AI-made content to carry a mark. New York now makes the advertiser disclose a synthetic performer it runs.
California’s AI Transparency Act took effect on 2 August 2026. Any generative system with more than a million users has to mark its output and embed a permanent hidden disclosure, and every breach carries a civil penalty. The duty to build the mark sits with the tool. The duty to use it, and the liability for running content without it, sits with the brand.
The law now lands on the advertiser
New York’s amended General Business Law is the clearest version of the shift. An advertiser that runs a synthetic performer it knows to be AI-made must disclose it, conspicuously, for the whole time it appears. The penalty rises on repeat. The obligation falls on the advertiser. The system that generated the face carries none of it.
California binds the tools to carry disclosure, down to a free public detector and a contract term that holds licensees to the same standard. New York binds the advertiser to use it. The two fit together into one exposure that lands on the brand.
The federal layer is coming behind them. The NO FAKES Act, through the Senate Judiciary Committee in June 2026, would give every person a property right over AI replicas of their voice and likeness. Anyone who distributes one without consent would carry civil liability. For a brand running synthetic creative, the exposure sits in the decision to publish the work.
Provenance has become the thing a brand pays for
As authentic and synthetic content become near indistinguishable, verifiable origin is what holds value. A brand that can prove where its creative came from can stand behind it. A brand that cannot is exposed the moment the work is questioned.
The industry has read this faster than the regulators. The ANA is building a tiered provenance standard for brand creative with hundreds of member companies, a way to grade how far the origin of a piece of work can be proved. The IAB’s 2026 transparency guidance recommends machine-readable provenance metadata, on the C2PA standard, inside programmatic advertising. Both treat origin as something a business now has to record and carry, not assume. The common thread is that provenance is becoming an asset a brand keeps, in the same way it keeps the rights to a logo or a campaign.
The business carries the content it cannot vouch for
The provenance gap is the distance between what a firm publishes and the origin it can stand behind. It widens every time synthetic output is published without a record of where it came from.
The gap is an enterprise exposure. A head of function cannot carry it, because it runs across legal, brand and financial risk at once. The disclosure duties are now live in law. The brand risk is running work it cannot verify. Penalties and infringement claims land straight on the balance sheet. None of it can be handed to the agency or the model maker, because whoever publishes owns what they publish.
The demand is already moving upstream, from the finished asset to the inputs that made it. A director of creative innovation at a VFX studio put it plainly.
“You have to track what prompts you’re using. You could get audited at the end of a production.”
Provenance has stopped being a label on the output. It is becoming the whole chain back to the input.
A firm that reads provenance as a production detail will keep treating it as the agency’s problem. The duty already sits at the top, and the exposure stays on the balance sheet whether or not it has been priced.
The sector is writing its own rules before the regulators finish
The ANA and IAB schemes are voluntary, and they are moving faster than statute. That is the signal. The market is pricing provenance before the law forces it, because the brands closest to the risk are not waiting to be told. When the biggest advertisers start grading their own creative for origin, they set a bar their agencies and their peers will have to meet.
The EU AI Act tightens the international frame. From August 2026 its Article 50 requires providers of systems that generate synthetic content to mark outputs as machine-readably AI-made. A creative and media company that operates across the US and Europe now carries two regimes at once, both live, neither optional.
A business that waits for the law to settle is already exposed under the rules that are running today. The settled version will only be stricter.
Three questions decide whether a company has priced the provenance gap.
- Can we vouch for the origin of everything we publish, or only for what we made ourselves?
- Where does the disclosure duty already bind us, and who inside the business owns it?
- If a synthetic asset we ran is challenged, who carries the liability, us or the agency?
The content a company cannot vouch for is the next thing it has to price.