Bloomberg Tech

Dan Neely on Why AI's Real Problem Is Permission at Scale

Dan Neely· Co-founder and CEO of Vermillio at Vermillio
·~5 min·English·Bloomberg
AI CompanyPolicyBusiness Strategy
TL;DR

Vermillio CEO Dan Neely on why "permission at scale" is the defining problem of AI and creative likeness, and how finding, removing, and licensing deepfakes hands control back to creators.

01The Distinction

Copyright vs. Likeness

Copyright protects what you made; likeness protects who you are — your voice, how you move, your name — and Neely says AI pushed that long-quiet right to the forefront.

Likeness is who you are. It's your voice, how you might move.

— Dan Neely, Bloomberg Tech
Key Insight
Copyright is the settled, well-understood right; likeness has existed just as long but stayed in the background. What's new isn't the right itself, it's that AI finally forced it to the front.
02The Scale Shift

From 19,000 to a Trillion

Neely says deepfake creations grew from about 19,000 in 2020 to about a trillion today.

Today, they're about a trillion. And so just to give you the kind of context of how much the scale has happened, it's just a gigantic problem.

— Dan Neely, Bloomberg Tech
Key Insight
At this scale, detection can't run as a human review queue — the practical response is automated identification across the whole internet, which is exactly the product Neely is selling.
03The Two-Step

Find It, Then Remove It or License It

Vermillio's technology finds unauthorized likeness across the internet, then either requests its removal or leaves it up to attribute and monetize.

First is let's identify all the unauthorized stuff. So our technology identifies across the entire Internet.

— Dan Neely, Bloomberg Tech
Key Insight
Making takedown requests and licensing two outputs of the same scan turns an adversarial legal process into a product decision: the detection that defends you is also the detection that can open a revenue stream.
04The Core Problem

The Messy Middle: Permission at Scale

Neely argues permission at scale is AI's real bottleneck — individual deals can't cover hundreds of thousands of people.

The entire infrastructure of AI, permission at scale is the problem because doing one on one deals is really not that complex, but they have to do it at a gigantic scale.

— Dan Neely, Bloomberg Tech
Key Insight
“The messy middle” names a transition, not a destination. The implied bet is that whoever builds the opt-in layer becomes the clearinghouse the whole industry has to route through.
05Where Value Accrues

Compensation Should Come From Compute

Neely predicts the larger payout will come from a cut of compute revenue, not ad revenue next to a clip.

The compute opportunity is actually the larger opportunity for these businesses and also for individual groups.

— Dan Neely, Bloomberg Tech
Key Insight
Pegging royalties to compute revenue rather than content views reframes likeness as an input to AI infrastructure, not a media asset — which would make creators stakeholders in model economics, not just licensors of clips.
06The Endgame

Optionality Is Mine

In Neely's telling, once a creator is protected the control flips back to them — the choice to opt in or out moves from the platforms to the individual.

It's not controlled by platforms. It's controlled by me as an individual.

— Dan Neely, Bloomberg Tech
Key Insight
The ordering is telling: in Neely's account creators come in concerned, get protected, and only then start weighing what else to do. Protection is the wedge; the licensing marketplace comes after it.