In every engagement
Everything we make can be
traced, disclosed and defended.
Provenance on generated assets. A record of what a human decided.
Consent architecture built for the enforcement dates that are actually coming.
the spine
You can’t bolt this
on afterwards.
Most studios treat governance as a line item inside a bigger engagement. That works when the output is made by people whose decisions are already documented in a chain of emails and version histories.
It stops working when the output is machine-generated at volume. By the time somebody asks where an asset came from, the answer either exists or it doesn’t — and it can’t be reconstructed.
So we built it the other way round. Governance is a property of how the work is done, not a service you buy separately. It’s in every engagement. You can extend it. You can’t remove it, and we’d rather lose a pitch than quote it out.
The three clocks
Three deadlines
your marketing is already inside.
Clock 01 India’s data protection regime
The rules are notified and the transition period is running. Supervision tightens through the back half of this year; full enforcement, with real adjudicatory powers, lands next year. Penalties reach into the hundreds of crores. What’s required includes valid consent, a system for governing it, and privacy notices in the languages your customers actually read.
Most organisations have the dates on a risk register and no plan underneath them.
Clock 02 Advertising disclosure for AI-generated content
India’s advertising standards body has published draft guidance, structured by risk. At the top, some uses are off the table regardless of how they’re labelled — generated endorsements, likenesses used without consent, claims that mislead. In the middle, disclosure is required where AI materially shapes a consumer’s decision. At the bottom, ordinary technical enhancement is exempt.
The final form isn’t settled. We build the classification step into production now, because retrofitting it across an existing library is expensive and retrofitting it across a live campaign is worse.
Clock 03 Transparency obligations for anyone selling into Europe
The headline deferral everyone heard about applies to high-risk systems. Marketing rarely is one. The transparency obligations that do apply to marketing output are already in force. If you sell into the EU, this one isn’t coming — it’s here.
[TO CONFIRM: add the specific enforcement dates for each of the three, verified on the day this page publishes. They move, and a wrong date on this page is worse than no date.]
the commercial argument
Disclosure is nearly free.
Concealment isn’t.
There’s a gap in this industry that’s worth knowing about, because it’s wide and it’s getting wider.
The overwhelming majority of advertising executives believe younger consumers feel positive about AI-generated advertising. Fewer than half of those consumers actually do. The industry is misreading its own audience, and the misreading has grown rather than shrunk.
But ask those same consumers about disclosure and the picture flips. A large majority say that knowing an ad was made with AI would either make them more likely to buy or make no difference at all.
So the one intervention that costs almost nothing is the one the industry is avoiding. We disclose.
What’s included
Built into
every engagement.
Provenance
chain
Content credentials attached to generated assets, a logged record of how each piece was made, and a chain you can hand to a legal team without a scramble.
Rights and
consent
Model access on commercially indemnified terms as standard. Consent chains for any talent or likeness used. Digital replica terms treated as a normal production item rather than an emergency.
Disclosure
logic
Risk classification as a workflow step rather than a final check. European transparency handling where you sell there. Synthetic performer disclosure for markets that now require it.
Data
governance
Consent architecture, notices in the languages you operate in, purpose limitation, retention — and data portability written into the contract rather than negotiated at exit.
The human-
choice record
Which decisions a person made, and when. It’s the operational form of the thing we say on our front page, and it’s auditable.
Principles
A note on
where this comes from.
India’s national guidance on AI governance sets out a handful of principles for how this technology should be built and used. Two of them are that people come first, and that systems should be understandable by design.
We started this studio on the line machines generate; humans choose. We didn’t take it from the guidance. We’re glad it agrees.
The Next Step
Find out what breaks when the rules
are applied to what
you’re already doing.
Governance exposure is one of the five dimensions in the Diagnostic. For most organisations it’s the one that produces the biggest surprise.