Red flags that make a data licensing introduction a bad idea

Which red flags stop a data licensing introduction?

Seven facts stop most introductions: the records belong to someone else, they are mainly consumer personal data or protected health information with no licensing basis, the archives are gone, a court or trustee controls the assets and has not been involved, the data is already licensed for AI training, the records were generated with AI to be sold, or nobody can export them. Add two baseline misses, a company that never reached 50+ full-time employees at peak (contractors excluded) and an owner who will not consider an exclusive license, and you have the full screen.

Each of these is far cheaper to find before the first email than after weeks of qualification. Screening first protects the owner's time, your standing with the owner and your credibility with SourceX.

You need only basic fit information to run it. Never ask an owner for samples, exports or screenshots; a partner's job ends at the introduction.

The pre-introduction red flag checklist

Work through the groups in order. A ticked box in any of the first four groups usually ends the introduction; the last group covers near misses, some of which change with time.

Who owns the records

• The most valuable records were created for clients and belong to them, as at many agencies, outsourcers and contact centers, and those clients have not consented.
• Core material came from contractors or vendors with no written assignment to the company.
• Client contracts limit the company's use of client information to delivering the engagement.

Whose information is in them

• The dataset is mostly consumer personal information, such as shoppers, app users or borrowers, and nothing in the privacy policy or terms supports licensing it.
• The records are mostly protected health information, such as medical records or claims, with no HIPAA authorization and no plan to de-identify.
• The company told customers in writing that their data would never be shared or used to train AI.

Whether the records still exist and can leave

• Archives were deleted, mailboxes purged or SaaS subscriptions cancelled without an export.
• No one at the company, and no vendor it trusts, can run exports from the systems that hold the history.
• Records were produced with AI tools for the purpose of selling them, rather than created by real work.

Who can sign, and what was signed before

• A bankruptcy court, trustee, receiver or assignee for the benefit of creditors controls the assets and has not been brought into the conversation.
• The same records have already been licensed for AI training.
• The owner has ruled out an exclusive AI-training license for an agreed term.

Baseline near misses

• Headcount never reached 50+ full-time employees at peak, contractors excluded.
• The company has only a short documented operating history.
• You have no route to the owner, CEO, CFO or another authorized sponsor.

The company fit checker runs a preliminary, non-binding version of the baseline questions with no contact details required.

Why does each red flag matter to AI buyers?

Buyers license records to train and evaluate AI systems, so they need a clean chain of rights, lawful content and a dataset that reflects real work. Each red flag breaks one of those three.

Three rows carry legal detail worth knowing. For health data, HHS guidance on the HIPAA de-identification standard describes two methods, Expert Determination and Safe Harbor, and information de-identified under either is no longer protected health information under the Privacy Rule. For consumer data, FTC staff have warned that quietly changing terms of service to allow AI training or third-party sharing may be unfair or deceptive, so a retroactive policy edit is not a fix. In bankruptcy, section 363 of the Bankruptcy Code limits sales of personally identifiable information that conflict with a privacy policy in effect when the case began, unless the court approves after a consumer privacy ombudsman is appointed. The guide to introductions during a restructuring covers who to involve.

This is general information, not legal, tax or financial advice. Confirm with the company's own counsel before relying on it.

How to read your results

Which warning signs are usually false alarms?

Some situations look like red flags but rarely are:

• The company was acquired or wound down. Operating, acquired and wound-down companies can all qualify if the data still exists.
• The history sits in retired systems. Archived systems often help, because long histories are what buyers value.
• The owner worries about who will see the data. Redaction and de-identification rules are settled before any work starts, and no record moves without an executed agreement and the company's authorization. The page on who sees data during a licensing deal explains the sequence.
• The owner fears reputational fallout. That is a judgment for the owner and any sponsor; the sponsor's view of reputational risk is a useful read before deciding.

How to tell an owner it is a no

A clean no keeps the relationship intact and leaves the door open if the facts change.

Next step

Screen your next candidate against the checklist above and the full who qualifies baseline, then register as a partner and introduce the companies that pass. Owners who would rather start themselves can apply at sourcex.si/apply.