What we will not broker, and why
We represent the company that owns the data, so before anything goes to an AI lab we check that the company can license it. These are the kinds of data we decline to take forward. Each comes with the reason a lab, a regulator or a court would stop it anyway, and what to do instead.
Data the company does not own
Scraped web content, purchased lists and records a customer or partner owns are not the company’s to license. A lab’s counsel asks where every dataset came from, and a claim from the real owner ends the deal.
Instead: License what the company created itself: its tickets, quotes, work orders and procedures.
Records under a contract that bars the use
Customer, client and partner agreements often limit how their information can be used or shared. If a contract does not allow de-identified use, licensing the records would breach it.
Instead: Read the agreements first. Records from customers whose terms allow it can go forward, and the rest stay out.
Data collected without a basis that covers this use
Privacy laws ask what people were told and agreed to when their information was collected. If that does not stretch to this use, cleaning it to the legal test is the only route, and some records cannot reach it.
Instead: Keep the records out, or have them de-identified to the test the law sets before any sample leaves.
Health, financial and student records that cannot be de-identified to the legal standard
The US health rule, the financial privacy rule and the student records rule each set a test for when data stops being personal. Free text in clinical notes and claim files often holds identifiers that survive a simple clean.
Instead: Use operational material such as scheduling, billing correspondence and procedures instead, and leave the sensitive files out.
Children’s data
Laws on children’s privacy limit what can be collected and released, and the identifiers in a child’s account, voice or location are personal information. The risk to the company is high and the risk to the child is higher.
Instead: Leave it out of any sample unless counsel confirms a lawful route.
Privileged and third-party confidential files
Legal privilege belongs to the client, not the firm, and a company holding another business’s trade secrets cannot license them. Neither can be cleaned into something the company may share.
Instead: License how the firm or company works, for example drafting histories without client confidences, and keep the confidential files out.
Data about a community that governs it
Some communities set rules for their own data, such as the First Nations principles of ownership, control, access and possession in Canada, and Indigenous data sovereignty principles in Australia. A company holding such records should have the community’s approval first.
Instead: Ask the community. If approval is given, the data can be considered with its terms attached.
Biometric identifiers collected without consent
Face templates, voiceprints and similar identifiers are treated as sensitive in several places, and recordings of staff, customers or bystanders can hold them.
Instead: Use recordings made with clear consent, and blur faces and remove other people’s voices before a sample leaves.
Data under legal hold or in a dispute
Records the company must preserve for litigation or an investigation, or whose ownership is disputed, cannot be handed on while the matter is open.
Instead: Wait until the hold ends and ownership is settled.
Anything the company has not approved
We show a lab only a sample the company has reviewed and approved, and the data does not leave until the company agrees. A sample the company has not seen is not one we can take to a lab.
Instead: Review the sample first. We work through the terms with you, and you make the final call.
General information, not legal advice: your counsel confirms what applies to your company. See the privacy rules by region and how we clean data in de-identification.
Questions about what we decline
What if my data includes some of these records?
Most company archives mix records. We separate them: the records the company owns and may license go forward, and the ones above stay out. A sample shows only what has passed the checks and your approval.
Who decides whether a record is in or out?
The company does, with its counsel. We apply the checks that the data is yours, that your customer contracts allow de-identified use and that the personal information can be removed, and we tell you what we find. Every license goes through counsel.
Can records on this list ever be licensed?
Sometimes, once the reason is dealt with: an owner agrees, a contract is amended, consent is obtained or the data is de-identified to the legal test. The rules differ by place, so see the privacy rules for your region.
Didn't find your question? Ask us directly.