Guides
AI data licensing providers compared: SF Data, Ooak Data, Telegraph Lab, Petrarch and IsarTrust
By Francesco Wiedemann · · 3 min read
In short
- A small group of companies licenses operational company data to AI labs: San Francisco Data, Ooak Data, Telegraph Lab, Petrarch and IsarTrust.
- They differ mainly in where raw data is processed, how anonymisation works, and what they publish about payouts.
- Published payout figures range from deals starting around 20,000 US dollars to target opportunities of up to 4 million US dollars.
- IsarTrust is the only provider in this overview based in Germany, with raw data processed exclusively in the EU.
This comparison was written by IsarTrust. It only uses information the providers publish on their own websites (as of 11 October 2026). If anything is out of date, please let us know at isartrust@knowlix.ai and we'll correct it.
Overview
| Provider | Based in | Focus | Anonymisation (as published) | Payouts (as published) |
|---|---|---|---|---|
| San Francisco Data | San Francisco, USA | Production traces and company data for AI labs | No public details | No public details |
| Ooak Data | Paris, France | Company data turned into RL environments for frontier labs | In-house engine; consistent replacement across sources; datasets delivered under a code name | "Up to $300K"; smallest closed deal $20K |
| Telegraph Lab | USA | Business data for AI training and evaluation | Sensitive fields removed or replaced with consistent placeholders | "$100K–$4M" target opportunities |
| Petrarch | USA | Enterprise operational records for training and benchmarks | In-house software (Lindo), runs on-premise at the supplier | No public details |
| IsarTrust | Munich, Germany | Operational data from European companies | Names replaced, details generalised or removed, result checked; processed in the EU | Licences from €20,000 |
San Francisco Data
San Francisco Data (sfdata.com) is based in San Francisco. Its public website currently centres on an essay about production traces as training data for AI. Further details on process, anonymisation or payouts haven't been published yet.
Ooak Data
Ooak Data describes itself as an applied AI research lab that turns real company data into reinforcement-learning environments. According to its supplier page, deals so far range from $20K to several hundred thousand dollars, raw archives are deleted within 30 days as a contract obligation, and buyers receive datasets under a code name. Ooak also works with companies that are winding down or have closed.
Telegraph Lab
Telegraph Lab licenses business data for AI training and evaluation and publishes "$100K–$4M" as target licensing opportunities, subject to data review. It shows example deals, offers a payout estimator and runs a referral and partner programme.
Petrarch
Petrarch sources operational records from businesses for AI training and benchmarks. Its de-identification software, Lindo, runs on-premise in the supplier's environment, so names, accounts and sensitive material are replaced before data reaches Petrarch. Access is by request.
IsarTrust
IsarTrust is based in Munich and is a product of Knowlix GmbH. It licenses data from European companies under its own licence agreements, anonymises it in the EU and licenses the refined dataset to AI developers (refine and license). Data processed on behalf of customers is excluded, works-council requirements are part of the process, and licences start at €20,000.
How to compare providers
- Location and law. Where is the provider based, and where is raw data processed?
- Anonymisation. Is it more than replacing names? Can you review a sample before delivery?
- Confidentiality. Is your company named to buyers, or is the dataset delivered under a code name?
- Role. Does the provider license the data itself, or does it connect you directly with buyers?
- Payment model. One-off, recurring, or both? How is the price determined?
- Exit. What happens to raw data after delivery, and can you end the licence?
More on the legal side in our GDPR guide for AI training data, and on the process in our guide to licensing company data.
Frequently asked questions
Which providers license company data to AI labs?
Among others San Francisco Data, Ooak Data, Telegraph Lab and Petrarch in the US and France, and IsarTrust in Germany.
Is there a European provider?
Ooak Data is based in Paris. IsarTrust is based in Munich and processes raw data exclusively in the EU.
How much do providers pay for company data?
Published figures range from deals starting around $20,000 to target opportunities of up to $4 million. The actual price depends on the data, its history, the tools covered and the rights granted.
Sources
Is your company's data licensable?
Three questions, about a minute, no data upload.
Check your eligibility →