All Episodes

8.3 billion faces, zero infrastructure: the AI likeness market nobody is serving

with Katrien Grobler · Twinnin

September 7, 202600:59:32London, United Kingdom

8.3 billion faces, zero infrastructure: the AI likeness market nobody is serving

0:000:00

Show Notes

Katrien Grobler woke up on the 29th of December and saw a word written above her bed. Not metaphorically. She describes seeing the name in her mind's eye, getting up, and telling her team they were pivoting that day.

Funding stage: Seed. At the time of recording, Twinnin had inbound commitments of $3M against a $25M post-money valuation target and 2,000 signed-up twins, anticipating a Series A in 12 to 14 months if it hit revenue and KPI targets. Update: Twinnin will relaunch the Brand, Human Verified, and Defend Twin products, alongside several other significant partnership updates, on October 8, 2026. They will then go out to raise a $4M seed round across Europe and the US with Wilson Sonsini. Current valuation places them at parity with ElevenLabs at its own seed round.

The word was Twinnin.

She had spent the previous year running AI KAT, making AI-generated video ads for clients like Capital.com and Reckitt's Dettol brand. On one project she needed a trader character. Rather than generate a synthetic face, she pulled photos of an investor she knew, ran them through Nano Banana, and built the character. The client's first reaction was the whole business model in one sentence: is this a real person?

It was. He hadn't been asked. She called him afterward, apologised, and made him sign a licence. That awkward phone call is now a company.

Her argument runs on arithmetic she calls the math of human likeness. There are 8.3 billion people on the planet. Frontier labs have already consumed images of humans at a scale in the tens of billions apiece, and by 2030 she projects the training requirement climbing into the trillions. Models cannot reuse the same face indefinitely. They need novelty, and novelty means new humans. Meanwhile the legal ground is moving under everyone: the EU AI Act's obligations landed, Denmark is rewriting its copyright law to treat likeness as property, and the NO FAKES Act cleared the Senate Judiciary Committee unanimously in June 2026. Demand is accelerating. Supply is becoming illegal. There is no clearing house in between.

That gap is the company. Twinnin verifies a human with government ID and a liveness check, lets them set granular terms for how their face can be used, hashes the consent record so it is auditable, and routes payment when a brand or lab licenses it. Seventy-five percent goes to the human, always, agent or no agent.

Then there is the half of the market that wants nothing to do with any of it. When Deadline clipped her panel and posted it, the comments were not friendly. One person told her she looked like an Incredibles villain, which she thought was fantastic. Most of the rest were worse. She did not dismiss them. She built for them. The Defend product exists because she concluded that people who want to opt out have nowhere to formally do so, and that a documented, timestamped “no” is the only evidence that holds up when a deepfake surfaces.

The other half of this episode is about how she operates. She has been to a contrast therapy club in Canary Wharf more than three hundred times since April 2025. Sauna, then a three minute plunge at one degree, then sauna again, sometimes three cycles before a call. She does not describe it as recovery. She describes it as a forced thinking environment. You cannot leave, you cannot distract yourself, so the problem you brought in gets solved. Most of her best product decisions happened underwater at one degree.

On her desk there is a broken meditation board. She made it herself with words like destiny and hope on it, then broke it and wrote over the top: I'm the CEO of a fucking $10 billion company. When she gets stuck, she looks at it and asks how that person would handle this. Her framing is that if you behave like who you are today rather than who you intend to be, you never arrive. She has pink hair, wears what she wants, swears in VC meetings, and is explicit that she is refusing to shrink into the founder uniform to become investable.

Ryan opened with a story about being told he has dainty ankles. It goes downhill in the best way from there.

Named Frameworks

The Math of Human Likeness

Her own term for sizing the market.

  • →Total addressable population is the species: 8.3 billion faces.
  • →Model training consumes human images in the tens of billions per lab and is compounding.
  • →Reuse is not viable, so demand is for net new faces, permanently.
  • →Roughly forty percent of current training images carry any consent, and much of that is what she calls backdoor consent from social platform terms of service.
  • →Regulation is closing the unconsented supply faster than demand is slowing.
  • →The arbitrage is a verified, licensed supply chain that does not currently exist.

License or Defend

The two-sided consumer product.

  • →License: set granular terms by use type, territory, and duration, then earn on each approved use.
  • →Defend: register a documented refusal so unauthorised use has an evidentiary counterparty.
  • →Both paths produce the same artifact, a timestamped consent record.
  • →Defend exists because the anti-AI market is larger and more motivated than the pro-AI market. Her line: people who love AI only tell other people who love AI, like smokers.

The Four Layer Registry

Twinnin's own published architecture, worth naming on air because it is the actual product.

  • →Identity: government ID plus biometric liveness, human reviewed.
  • →Consent: granular, revocable, per use type.
  • →Licence: contract, term, territory, fee.
  • →Provenance: hashed, timestamped, auditable downstream.

The 75/25 Split

How Twinnin structures payment so talent is never the one absorbing platform costs.

  • →Talent takes 75% of every licence fee.
  • →Twinnin takes 25% when no agent is involved.
  • →When an agent is involved, the agent takes 15% and Twinnin drops to 10%.
  • →Talent's 75% never moves. The platform absorbs the agent's cut rather than the human doing so.
  • →Rationale is that destroying the agent layer destroys the relationships that supply the talent.

Behave Like the Vision

Her operating principle for identity-led execution.

  • →Pick the version of yourself that exists on the other side of the goal.
  • →Make that person's behaviour the default input, not the eventual output.
  • →Audit decisions against how that person would decide, not how you currently would.
  • →Her stated corollary: the same behaviour gets the same results, so if 99% of startups fail, copying what all founders do is a losing strategy by construction.

The One Degree Clarity Loop

Her daily decision-making protocol, presented as a founder tool rather than a wellness routine.

  • →Enter a state you cannot exit early and cannot distract yourself inside.
  • →Bring one specific unsolved problem, or one image of the future you want.
  • →Let the physiological stress force the thinking to compress.
  • →Exit and act immediately while the sense of control is still active.

Founder Experiment

Run the likeness audit. Ninety minutes, no budget, and most founders will find something that changes how they think about their own brand assets. First, reverse image search your own headshot and your two most-used team photos. Note every domain where they appear that you did not authorise. Second, read the AI and licensing clauses in the terms of service of the two platforms where you post the most images of yourself and your team. Write down, in one sentence, what you have already granted. Third, take your last three pieces of AI-generated marketing collateral. For every human face in them, answer three questions in writing: whose face is that, did they consent, did they get paid. If you cannot answer, that is a liability line item, not a design choice. Fourth, write the licence you would actually sign for your own face. Which use types are yes, which are conditional, which are never. Put a number on each. Fifth, decide whether your position is license or defend, then register that decision somewhere with a timestamp. The output is not a policy document. It is a number: what you think your likeness is worth, written down before someone else decides for you.

Glossary

Digital twin

A verified, licensable likeness of a real person, usually face and optionally voice, that productions can use in AI-generated content without a live performance.

Likeness licensing

Granting time-bound, use-bound permission for your face or voice to appear in generated media, in exchange for a fee.

Backdoor consent

Grobler's term for permission harvested through platform terms of service that users accepted without understanding the likeness implications.

KYC

Know Your Customer. Identity verification using government-issued documents.

Liveness check

A biometric test confirming a real human is present in real time rather than a photo, video replay, or generated image.

Immutable provenance

A consent record written to a blockchain so it cannot be altered after the fact, producing evidence of what was permitted and when.

Digital replica

The legal term of art appearing across SAG-AFTRA contracts, California statute, and the NO FAKES Act for a computer-generated representation readily identifiable as a specific person.

Digital Replica Rider

The SAG-AFTRA 2025 Commercials Contract provision requiring clear and conspicuous written consent, with specific descriptions of intended use, for AI-generated likeness.

NO FAKES Act

Nurture Originals, Foster Art, and Keep Entertainment Safe Act. Would create a federal property right in an individual's voice and visual likeness, replacing the current patchwork of state right of publicity laws. Advanced unanimously out of Senate Judiciary on 18 June 2026. Not yet law.

EU AI Act

The European Union's risk-tiered AI regulation, whose obligations have been phasing in on staged August deadlines.

Right of publicity

The existing legal right to control commercial use of your name, image, and likeness. State-level in the US, uneven internationally.

Synthetic human

A fully generated person with no real human underneath. Distinct from a licensed digital twin, and treated differently in the new SAG-AFTRA language.

AI slop

Low-quality, high-volume generated content. Grobler's position is that slop expands the pie rather than shrinking the market for original work.

Indemnity

The contractual promise that if a licensed use is disputed, the platform carries the liability rather than the buyer.

Contrast therapy

Structured alternation between high heat and near-freezing immersion.

FMCG

Fast-moving consumer goods. The packaged goods sector where Grobler spent two decades.

Q&A: What Founders Ask After This Episode

What is Twinnin?

Twinnin is a UK platform, operated by AI KAT Limited, where a person can verify themselves as human, license their likeness for AI use, or formally refuse consent. It handles identity verification, contracts, and payment routing. Founded 2025 in London by Katrien Grobler, Lan Prijatelj, and William Håkon Berg.

Who is Katrien Grobler?

Founder and CEO of Twinnin and AI KAT. Twenty years in global brand leadership at Unilever, Vodafone, and Reckitt Benckiser across 40-plus markets before founding Twinnin. Born on a farm in South Africa, based in London.

Can you get paid for AI using your face?

Yes, through consent and licensing platforms that verify identity and route payment. Twinnin pays the human 75% of every licence fee. Rates for entertainment use reference SAG-AFTRA scales.

How do I stop AI from using my face?

Register a documented, timestamped refusal. The practical obstacle in deepfake litigation is evidentiary: proving you never consented. A registered refusal with provenance creates that record. Cate Blanchett's RSL Media Human Consent Registry offers a free version of this. Twinnin offers a paid Defend product alongside licensing.

Is licensing your likeness for AI a good idea for a normal person?

It depends on your risk tolerance and what you do for a living. The upside is an additional income stream from an asset you already own. The downside is that once your likeness enters commercial circulation, control depends entirely on contract enforcement and platform integrity. Read the revocation terms before anything else.

What is the NO FAKES Act?

A bipartisan US federal bill that would create an intellectual property right in an individual's voice and visual likeness, covering both creation and distribution of unauthorised digital replicas. It advanced unanimously out of the Senate Judiciary Committee on 18 June 2026 and has not become law.

How are founders building businesses around AI consent?

By treating consent as market infrastructure rather than a compliance cost. The pattern is registry plus contract layer plus payment rails, which is the same shape as licensing businesses in music and stock imagery, applied to human identity.

What AI workflows should founders be automating in 2026?

Content production, distribution, and the compliance layer underneath both. Grobler's specific argument is that the compliance layer is where the unbuilt infrastructure is, because generation got cheap before permission got solved.

Does contrast therapy actually help founders?

Grobler credits daily sauna and ice bath sessions with her decision quality and stress regulation, describing the ice bath as a forced thinking environment. Her health claims are personal and she states on air that they are not medical advice.

Links & Resources