Founded by former leaders at Scale AI and Meta Superintelligence Labs.
Longitude Labs supplies training and evaluation data to frontier AI laboratories.
Customers
Companies in the Mag 7 · Multiple top-5 frontier labs · Fortune 500 enterprises
Backed by
We license the documents behind investment decisions.
Purpose
Longitude Labs acquires text-based financial documents, including investment materials, teasers, CIMs, process materials and the supporting legal documents, and uses them as a basis for AI model research and training.
What we buy
Documents that contain substantive analysis, ideally full M&A dealrooms, including IC memos, CIMs, diligence reports, lender presentations, investor letters and the related deal-folder material.
Why partners partner with us
- Compensation for work that has already been completed
- Best-in-class anonymization
- A contribution to the development of AI models
Four kinds of financial documents qualify.
1Deal & process materials
- CIMs / CIPs, authored or received on past processes, incl. RE OMs
- Management presentations
- Process letters, teasers & bid instructions (ride-along)
- Deal-folder ride-alongs: term sheets, LOIs, fee letters, NDAs
- Fairness opinions & board decks
2Investment committee & research
- IC memos & IC decks (first-round or final, including iterations)
- Sector & industry deep-dives, landscape & market-sizing reports
- Earnings reviews & thesis updates with real analysis
- Valuation reports & fairness analyses
3Credit & financing
- Lender / financing presentations under 50 pages, authored or received
- Rating-agency decks under 50 pages
- Credit memos & credit-committee papers
- Financing & credit agreements, debt term sheets & commitment letters
4Fund & diligence materials
- QoE reports, diligence reports & financial DD summaries under 50 pages
- Commercial & market due-diligence studies
- LP updates, quarterly / annual investor letters & annual meeting decks
- PPMs & offering memoranda, LPAs & side letters under 50 pages
- Fund marketing decks / pitchbooks & completed DDQs
Not an exhaustive list. If a document carries real financial analysis, send it through. Out of scope: thin content, anything publicly available, AI-generated content, non-English language content, blank forms.
Every document is anonymized, audited and never published.
Anonymization process
1Multi-step anonymization process
Software built and maintained in-house removes all identifiers, including client, counterparty and personnel identifiers, deal code names, logos, document footers and visual identifiers
2Residual-PII audit
After anonymization, each page is checked again for text and visual residuals. Findings return as an audit note and are corrected if possible
3In-house review
Every anonymization is checked by full-time professionals
4Accepted & licensed
Only documents that pass are licensed. Around 20% of what we acquire fails the check and is never used
How materials are handled
1Controlled systems
Materials are only reviewed on systems managed by Longitude Labs
2Need-to-know access
Access is limited to personnel who require it for review
3Anonymization
Partners send originals because anonymization at this standard cannot be done in-house
4Cold storage
Materials are put in cold storage within 90 days to minimize cybersecurity risks
5No publication
Documents are never published and accepted materials are used only as a basis for AI research and training
We agree on the price before any file is sent.
- 1
NDA
We sign a mutual NDA before anything is shared, covering the samples and everything that follows
- 2
Samples
You share 2–3 sample documents, anonymized on request, so we can see what the collection looks like
Your step - 3
Fit check
We assess whether the material is what we are looking for and give you feedback
- 4
Sizing
You give us a view of the collection (file count and types) via screenshots, a short walkthrough or the inventory prompt
Your step - 5
Quote
We quote the total for the whole archive. Nothing beyond the samples has left your machine at this point
- 6
Agreement
Once you accept the price, we sign the licensing agreement, and you send the files via Google Drive or box.com
- 7
Payout
Payment by wire, usually within 10 days of the intro call
Your side
1–2 hours in total, mostly in steps 2 & 4
Our side
Assessment, quote, anonymization and licensing, free of cost
Track record
c. $1m paid out to partners in August and September 2026 alone
70+ data partnerships signed
Size your archive with Claude or ChatGPT, on your own machine.
Claude
- Download the Claude desktop app
- Open a new session and point Claude Cowork at the folder of your file archive, using “Project or folder” > “Add folder”
- Paste the prompt
ChatGPT
- On the homepage, switch the toggle at the top to Work
- Create a new project and add the folders holding your archive
- Paste the prompt with the project selected
This will serve as the basis for our quote. The inventory is compiled locally from filenames and folder structure. Only the table and the description are shared with us.
Prompt
This folder contains a full document drive that we want to analyze. Go through every subfolder and build an inventory using 9 distinct categories that make sense to you. Output one table with each category as a row. As columns, give me: column 1 - total number of documents / column 2 - # of word/pdf documents / column 3 - # of xls documents / column 4 - # of ppt/pptx documents / column 5 - # of other file type documents / column 6 - total number of pages / column 7 - show 2-3 typical, anonymized filenames for spot-checking. Anything substantive that doesn't fit goes under the 10th category called "Other" with a one-line description. List number of dealrooms with buyer thinking (contains IC documents, financial model) and without buyer thinking. Include a second table with document age distribution across all dealrooms. Give me the total sum / number of docs and pages in a final row. Give me the total file size of all documents combined. Never mention client names or PII in the analysis. Eventually write a 5-10 bullet points as description of the file archive. Final output as Word docx.
Frequently asked questions.
1Do I have to share the full dataset before I know the price?
No. Determining the quote for the full dataset only requires running a Claude/ChatGPT prompt through the folder. The result will not include any confidential information. If you don’t feel comfortable running Claude/ChatGPT on your hard drive, please contact us to find an alternative way of quoting your archive.
2Can I trust you?
We have raised $50m+ from Tier 1 investors such as Andreessen Horowitz and Greylock. We consider ourselves best-in-class when it comes to anonymization and confidentiality, as it is key to our success. Besides your data, we also keep client names strictly confidential, similar to a law firm.
3Will our firm or its clients be identifiable?
No. All PII, including client and counterparty names, matter numbers, deal code names, bank details, company registry numbers, named facilities, visual identifiers, digital footprints and similarly identifiable information, will be thoroughly removed. Our anonymization software was developed as a result of trying out many solutions on the market, all of which didn’t meet our quality standards.
4What happens to the documents?
Accepted materials are used for research and training. Documents are never published. We use anonymized versions internally to create AI training tasks and may share some anonymized versions with our partners (frontier labs).
5Can I see the anonymized versions of my documents?
Gladly. We send examples of anonymized documents on request, and we can put you in touch with past partners.
6Which compensation is typical?
The highest compensation we have paid out to date is $500,000 for a single partnership. The smallest partnership was $6,000.
Existing partners report timely payment.
“We’ve been working with Meridian1 for almost a year. What sounded like an unconventional offer turned out to be a seamless experience. Payment is always timely, and the team is a pleasure to work with.”
August 2026c. $290kPayout to date
“Meridian1 is a great partner to do business with. They always respond almost instantly and keep their promises.”
August 2026c. $24kPayout to date
We do not disclose the identity of our partners. On request we introduce prospective partners to existing partners who have agreed to act as references.
1 Testimonials predate the renaming from Meridian to Longitude Labs.
