01
Introductory discovery
Discuss systems, operating history, and goals. No file uploads, raw exports, or logins are requested.
Business data licensing · Privacy & Control
Explore the opportunity.
Stay in control.
What happens to your company's information—from the first conversation to a possible licensing agreement? Here's a practical guide to the described process and the decisions that remain yours.
No raw files during discovery
Written approval before transfer
De-identification required before use
From discovery to delivery
Who sees what—and when?
This summarizes the process described in the October 22, 2026 privacy guide. Specific protections must be verified in the actual agreements for your transaction.
02
Representation agreement
A separate agreement defines how the opportunity will be represented; signing it does not itself deliver data.
03
Data manifest
A description of the systems, approximate volumes, and years of records is assembled. The intermediary receives information about the data, not raw datasets.
04
Anonymous buyer profile
Prospective vetted AI labs review a profile. The company's identity is disclosed to a buyer only with the company's approval.
05
Offers and written approval
The owner chooses whether to accept any offer, approves the buyer and permitted scope, and decides what is excluded. No raw data moves at this stage.
06
Direct, limited delivery
If a deal proceeds, the company transfers only the approved export directly to the AI lab using an encrypted, time-limited channel. The intermediary does not receive raw data or credentials.
07
De-identification before use
Inside the lab's secure environment, identifying information is removed or transformed under the license before the data is used. The lab may receive the approved export before this step.
08
License end
Contractual deletion or return terms apply to the received data. Models already trained may remain in use; training effects generally cannot be extracted from a model afterward.
Safeguards explained
Two ways personal information is protected.
Redaction · Remove it
Data the buyer does not need can be withheld entirely. For example, an unnecessary identification-number column can be deleted or an entire HR system excluded.
Example: “Call Maria about invoice 8124” → “Call [REDACTED] about [REDACTED].”
Transformation · Replace it
Names, email addresses, dates, and other details can be replaced or generalized so relationships and workflows remain useful without retaining the original identifiers.
Example: a person's name becomes a consistent stand-in such as Person_0412.
Why the distinction matters: Redaction removes information, while transformation preserves useful patterns. The exact approach and its privacy risks depend on the data, the lab's controls, and the contract.
You control
- Which systems, years, fields, and records are excluded.
- Which buyer and permitted uses you approve.
- Review of samples before delivery.
- Whether to accept or decline an offer.
- Review of your contracts, policies, and legal rights with counsel.
License obligations described
- De-identify information before use.
- Restrict uses and prohibit re-identification.
- Keep information encrypted and access limited.
- Do not resell without consent.
- Follow contractual incident notice and deletion or return terms.
Information excluded from licensing
According to the guide: information involving minors; classified or export-controlled material; passwords and credentials; card and bank account numbers; health and biometric information; privileged legal materials; and client or third-party content that the company has no right to license.
Frequently asked questions
Questions worth asking before you proceed.
Do you need my customer files or system passwords for a call?
No. Discovery only requires a general discussion. Do not send files, exports, logins, or credentials to me.
Will an AI lab see identifying information?
The guide describes a direct company-to-lab transfer of an approved export, followed by lab-side de-identification before use. That means raw identifying details may be present in the lab's secure environment before processing. Ask to review the exact safeguards and allowed uses in writing.
Can I exclude a department, folder, or time period?
Yes, the described process permits exclusions of systems, fields, records, and years before any export. The approved scope must be reflected in the agreement.
Is de-identification a guarantee of zero privacy risk?
No. Privacy risk depends on the information, techniques, access controls, contractual protections, and applicable laws. Obtain independent legal and privacy advice before licensing data.
Can trained AI models be undone after a license ends?
Generally no. The guide states that models already trained usually remain in use even when the underlying received data is subject to deletion or return terms. This is why permitted uses and duration should be agreed before transfer.
Talk through your questions
Explore whether your company could qualify.
A short introductory call does not require confidential files or a commitment to license data.
Schedule a Quick Chat →This page summarizes a partner privacy guide dated October 22, 2026; it is informational, not legal advice or a certification of compliance. Individual transaction terms may differ. Independent counsel should review data rights, privacy policies, contracts, and the proposed license.