Can you license lab notebooks and experiment records to AI companies?
A published paper shows what worked. A lab notebook shows everything else: the settings tried, the result that was wrong, the reason it was wrong and what changed next. Experiment records tied to their measured outcomes are scarce because most labs never published the failures. A lab that kept them has something rare, and several questions to settle first.
What a lab archive contains
| Notebook entries with the reasoning | What was tried, why, and what the researcher concluded, in the order it happened. |
| Structured run records | Inputs and settings linked to measured results, in an electronic system or a spreadsheet. |
| Failed and repeated experiments | Runs that did not work and runs repeated on another day, with the notes on why. |
| Analysis files and instrument output | The data from the instrument and the analysis that interpreted it. |
| Protocols and method changes | How a procedure changed over time, with the reason for each change. |
Failures are as useful as successes
A record that includes only successful experiments teaches a model that everything works. A record with the failures, the repeats and the conditions that changed shows how results depend on settings. That is a hard thing to learn from the literature.
The link between settings and results is the center of it. Notes without outcomes, or outcomes without settings, are worth far less than the pair.
Ownership, publication and consent
Records may belong to the company, to a sponsor who paid for the work, to a university partner or to a collaborator. Each of these can have rights, so read the research agreements first.
Data that has already been published is less interesting. Anything involving patients, trial participants or genomic material carries consent and regulatory questions, and should be taken to counsel before any preparation.
What stays with sponsors and patients
| Sponsor-owned and partner-owned data | Left out unless the agreement gives the company the right to share it. |
| Patient and participant information | Excluded unless it can be fully removed and the consent allows it. |
| Unpublished results the company wants to protect | Left out at the owner’s choice, including anything tied to a pending patent. |
An example, invented for illustration
A small materials company that wound down kept its electronic notebook and its run records, several years deep, with every build logged and the failures noted. A set includes runs the company owns outright, links each notebook entry to its measurements and replaces researcher names with stand-ins. Work done for a sponsor is left out, and the former owners approve a sample before anything is exported.
Questions owners ask about this
Our results were never published. Does that help or hurt?
Unpublished results are more interesting, since they cannot be found elsewhere. Check patent and sponsor terms first, because those can limit what you may share.
Our lab has closed. Can its records still be licensed?
Often, yes, if someone has the authority to sign and the rights are clear. Companies that have wound down can be a good fit, and the paperwork about who owns what comes first.
Is a spreadsheet enough or do we need a notebook system?
A spreadsheet can work if it links settings to results. The format matters less than whether the pairing is complete and consistent.
Didn't find your question? Ask us directly.
Keep reading
Related kinds of records
- Project and task history: Jira, Asana, Linear, Monday, Trello, ClickUp and Airtable comments and status changes, not task titles.
- SOPs and knowledge bases: Procedures with their edit history and comments; published help articles add nothing.
- Inspection reports and photos: A finding beside its photos and its fix, with faces, plates and addresses removed first.