Training data provenance audit for AI startup sales
A fixed-fee audit that flags every licence encumbrance in training data, delivered as one signed document a UK AI startup hands to buyers.
You get the checking already done: the evidence behind the idea, the sources it came from, and the objections it survived, all open below so you can judge them yourself.
UK·41 sources·8/8 checks·Verified 2 days ago
Sourced from aiexponent.com, wilmerhale.com, aetos-data.com
6 in 100or fewer get through
This is one of them. 1,444 ideas went through the filter and 94.5% of them died on cited evidence. Read what killed them.
One page of the evidence record142717e797740247
Enterprise buyers also add AI governance sections to standard vendor risk assessments covering model provenance and training data rights, referencing the NIST AI Risk Management Framework and ISO 42001.
The workings are the fourth document, and its arithmetic is computed rather than written.
The numbers are in the pack. Pricing mechanics and unit economics, the assumptions behind them listed in full, and anything the research could not stand up is marked so.
Secure checkout via Stripe. A pack is evidence-backed research, not a promise of business success. See our refund policy.
What’s inside your pack
The moment you pay, you download the whole pack. 14 documents, no drip feed, no login.
Inside the pack14 documents
Where this starts
The opportunity on one page: what it is, what checked out, and what we do not claim.
What you would be selling
The thing that changes hands for money, written as an offer a buyer could accept today.
The field: who is already there
The people already selling into this, what they charge, and where they leave the door open.
The numbers
Pricing and the numbers behind it. Anything we could not verify is marked missing, never made up.
What would sink this
The case against, at full strength, so you meet the objection here rather than from a customer.
What you build
What to build, in what order, on what stack. Includes the non-goals for v1 and what would kill this.
How the first customers find you
Where your first customers come from. Named channels, the beachhead to start in, and the signals that say stop.
How it runs once it works
How it runs once someone pays. Delivery, capacity limits, the compliance you cannot skip, and where the manual work sits.
Your first fortnight
Six steps for days one to seven: confirm the buyer, size the smallest paid offer, pick one channel.
The toolkit
The named tools, services and suppliers this runs on, with what each one is actually for.
Copy you can paste
Launch copy you can send today: listing page, outreach, social. Claim-checked like the research.
How to know in 30 days
The one month test: what to run, what number to watch, and the reading that says walk away.
Everything we read, once
Each check, what it found, and the source behind it, without hunting through the plans.
Every check, in full41 sources
Every check this pack faced, each verdict, and a clickable source behind every claim.
what lands in your folder
The pack, readableindex.html
Open this first and read the whole pack in order in your browser. No install, no account.
The pack, typeset for printComplete_Pack.pdf
Every document in one printable PDF. Open it on a phone, or put it in front of someone.
Your first fortnight, on one pageFirst_Fortnight.html
A single sheet to print and work from. Days one to fourteen, nothing else on it.
Every assumption, as a spreadsheetAssumptions.csv
Every number the financial model rests on, in a file your spreadsheet opens directly.
Marketing copy, ready to pasteMarketing_Assets.txt
The launch copy as plain text, so it goes straight into an email, a listing or a post.
The machine-readable recordmanifest.jsonld
What this pack is, in structured data, so a tool can read it as easily as you can.
A web page you can read, a PDF you can print and a spreadsheet you can open. Yours to keep, edit, or paste anywhere. No login, no subscription.
Instant download
Could you run this?
Behind the research is a business somebody has to actually operate. Here is who they would be selling to, and how soon the first money arrives.
Sells to businesses
Part automated
Evenings-friendly
Vertical tool
Market
UKThe opportunity, its evidence and its economics are researched for this market. The pack itself is priced and sold in GBP.
Who pays
Early stage UK AI startup CTOs and founder-operators preparing for enterprise procurement, technical due diligence, or AI liability insurance underwriting. Reached by direct outreach to UK AI founders.
How we tried to kill it
Each check is an attack, not a rubber stamp. An idea dies on the first check where cited evidence goes against it. This one survived all 8. Finding nothing is not the same as finding a green light; see how each check works on /how-it-works.
The 6 below run on every idea, whatever it is. This one's lane ran 2 more on top, for 8 in all.
Exactly what this pack covers, plus a blurred look at the document you receive.
Training data provenance audit for AI startup sales
01 · A cross-reference process for manifests against public licence sources, opt-out lists, and published takedown notices
The EU AI Act's obligations for general-purpose AI models applied from 2 August 2025, and include a copyright policy and a publicly available summary of training content using the AI Office template.
The European Commission published that mandatory template on 24 July 2025.
Month 1 revenue
£2,500
LTV : CAC
12.5×
Payback
1 month
Unlocks the moment you buy
01A cross-reference process for manifests against public licence sources, opt-out lists, and published takedown notices
A look inside
Real, sourced lines taken straight from the pack. Every source below is a live link: open one and check the claim before you buy.
Extract · evidence record
Enterprise buyers also add AI governance sections to standard vendor risk assessments covering model provenance and training data rights, referencing the NIST AI Risk Management Framework and ISO 42001.
The EU AI Act's obligations for general-purpose AI models applied from 2 August 2025, and include a copyright policy and a publicly available summary of training content using the AI Office template.
The workings are the fourth document, and its arithmetic is computed rather than written.
The numbers are in the pack. Pricing mechanics and unit economics, the assumptions behind them listed in full, and anything the research could not stand up is marked so.
Secure checkout via Stripe. A pack is evidence-backed research, not a promise of business success. See our refund policy.