Is licensing business data safe?

Start without uploading records

Maryah onboarding uses answers, counts and screenshots first. Keep raw records in your own systems until a buyer deal is agreed. Screenshots should show the requested counts, not expose unnecessary names, message contents or credentials.

Replacing a name is one step, not the whole job

In a made-up example, "Jim Smith resolved a delivery issue" becomes "Person 14 resolved a delivery issue." The same label can keep the work connected across records. But a job title, a rare incident or other details could still identify someone. Review needs to cover context and free text, not just a list of names. The UK Information Commissioner's Office distinguishes pseudonymisation from anonymisation. Replacing identifiers can reduce risk, but does not automatically take data outside data protection law.

Agree what never enters the process

Scope out passwords, API keys, financial account details, government IDs, health records and children's data. You can also exclude systems, folders and commercially sensitive material. Ask how exclusions are checked in attachments and free-text fields as well as structured columns.

Know where data goes and who handles it

Before transfer, establish the destination, who can access it, who prepares it and what records show the work was completed. The partner walkthrough describes preparation by a de-identification partner or by the buyer under contract in a controlled environment. Do not assume every deal uses an identical system or that a marketing label proves its security.

Keep ownership and usage rights separate

Licensing is permission to use agreed records, not a transfer of your company. Review ownership, the licensed scope, seller exclusivity, buyer restrictions and how future sales work. Do not assume you approve every transaction separately: confirm the actual approval rules in your agreement.

What can buyers do?

The licence may permit training AI, fine-tuning an existing model and creating synthetic training examples. Raw-data redistribution, re-identification and other uses need explicit restrictions. Your counsel should check the actual agreement and your authority to license the underlying records.

What happens when a licence ends?

The terms may require deletion of the raw dataset and allow future licensing to stop. That does not necessarily remove models, outputs or other derivatives already created. Deleting files is not the same as undoing model training. Understand what survives before agreeing, rather than expecting a complete reversal later.

Questions to take into the call

What is excluded? Who prepares the records? Where are they processed? What evidence documents preparation? Which buyers and uses are restricted? What can stop, and what survives termination?

This guide explains the process. It does not replace legal advice or a compliance review. Requirements depend on your records, contracts and the relevant jurisdictions. Before sharing personal or confidential information, involve your privacy or legal adviser.