DillisData Get my estimate

Updated

For biotech and life science labs

Sell Biotech Data to AI Companies

Short answer

Biotech and life sciences labs can license experiment records they own and have not published. Some AI data programs ask for electronic lab notebooks or structured records, with 1,000 or more runs that show inputs and measured outcomes. Failed runs count. Data a sponsor or partner owns does not.

To sell biotech data to AI companies, most labs grant a license. The license lets a program work from a copy of your experiment records. Your lab keeps the originals and its ownership. Our general guide explains how companies sell data to AI companies in other fields.

What sets a lab apart is the work that never made it into a paper. Some AI data programs want unpublished runs, and failed runs above all. A journal article shows the win. Your notebook shows the many tries before it, and that is hard to find anywhere else.

DillisData is a data agency that matches labs with programs that fit. The signing happens between you and the program, and your records go from your systems to theirs. We never hold any of it. We send you no invoice, and programs pay us a referral fee.

  • Failed runs count
  • Ownership checked first
  • Records bypass DillisData

See what your data is worth

Progress

What programs look for

What do AI companies look for in biotech labs?

Program rules differ, and they can change. Some programs work with labs of about 10 to 1,000 employees, anywhere, that hold several years of experiment archives. Here is what they ask for, and the reason.

What some programs ask forMust have or bonusWhy they want it
Electronic lab notebooks, or ELNs, or structured experiment records your organization ownsMust haveDated, structured records show what was tried and in what order.
1,000 or more runs with their inputs and measured outcomesMust haveA model can link what went in to what came out, across many tries.
Failed runs kept alongside the good onesMust havePapers rarely report failures, so this is the rare part.
Data that hasn't been publishedMust havePublished work is already out in the world. A program can't get yours elsewhere.
Clinical study reports with patient-level data, tables and analysis plansBonusThey show how a trial was planned, run and read.
Genomic data collected with research consentBonusPrograms look for the consent as much as the data.
In-house assay and analysis toolsBonusAn assay is a lab test. Your own tools show how your team measures things.
Scientists' write-ups of their experimentsBonusWrite-ups capture the reasoning behind each choice.

Who is not a fit

Which labs and records are not a fit?

Ownership trips up more labs than anything else here. Go through this list before you count your runs. Whatever you exclude gets written into the contract.

Data owned by a sponsor or partner

Work done for a sponsor, or under a collaboration, may belong to someone else. CROs, the contract labs that run studies for clients, need to read this one twice. Read each agreement and any material transfer terms.

Data already published

Some programs rule this out. Your unpublished runs behind a paper may still count, so sort them apart.

Device or AI software companies

If your data is mainly software logs, some programs say this is not a fit. Run records from real experiments are the point.

Export-controlled and defense work

Some programs turn this away. The US Commerce Control List has a category for chemicals, microorganisms and toxins. Ask your lawyer whether your work touches it.

Grant-funded work with strings attached

Grants and government contracts can carry their own terms on data and inventions. Check them before you count those runs as yours.

Other companies some programs skip

Some general programs also pass on staffing agencies and cannabis firms, plus franchisees without control of their own systems.

Ownership and consent

Can a biotech company license data it doesn't fully own?

Generally, it can license only what it has the right to license. So start with a simple sort. Put your runs into three piles: clearly yours, clearly someone else's, and unsure. Sponsored research agreements, collaborations, licenses from universities and past deals all belong in that review. Your lawyer should look at the unsure pile. Clean ownership matters to programs.

Keep an eye on what the records give away. A license hands over a copy of your methods and results. So think about your trade secrets, and about any invention you haven't filed on yet. Ask your patent lawyer before those records go in. You choose what is in and what stays out, and that gets written down.

Human data raises a second set of questions. Research consent forms often say whether data may be reused once identifiers come off (45 CFR 46.116). Whether that wording reaches a paid AI license is a question for your lawyer and your review board. If a hospital or clinic gave you patient data, its contract may also limit reuse.

HIPAA may apply as well, if the records came from a covered provider or health plan. HIPAA's test for de-identified data has two methods (45 CFR 164.514), which HHS explains here. Trial record numbers can count as identifiers under Safe Harbor. Ask a qualified expert how this applies to genomic data, and what to do about it.

Good lab habits can help you here. FDA's lab rules for nonclinical studies require keeping raw data and protocols in an orderly archive (21 CFR 58.190). A lab that follows them may already have much of this in shape. This is general information, not legal advice.

See what your experiment archive could bring

Programs often offer $100K to $5M+, largely depending on your lab's size. In a lab, volume and depth of records matter as well. That is a range and never a promise. The estimate gives you a figure for your lab.

Estimate my lab's records

Where it lives

Where do your lab records live?

Most labs keep this across several tools, plus old folders nobody has opened in years. These are common ones. Yours may be different.

RecordWhere it usually livesWhat to check
Experiment notes and protocolsELNs like Benchling, LabArchives or SciNoteCan you export with dates and authors kept?
Sample and test resultsLIMS, the lab's sample tracking system, like LabWareDo results link back to the run that made them?
Research data across teamsPlatforms like Dotmatics or IDBSIs the same experiment split across tools?
Raw instrument filesInstrument computers and network drivesAre failed runs still there, or were they purged?
Trial documentsClinical folders and vendor systemsWho owns them, you or a sponsor?
Write-ups and decisionsShared drives, slides, email and chatDo they explain why a project moved on or stopped?

How it works

How does a lab get from archive to signed license?

Your lawyer's ownership review comes before anything else moves. We stay out of your files throughout.

01

Take the quiz

Tell us your size, your systems and how many years of records you hold. There is no upload and no login.

2 minutes
02

Review your matches

See your estimate and the programs that fit, plus why each one does. The estimate is not an offer.

03

Sort what you own

Your lawyer and lab lead sort runs by ownership, consent and publication status. Then you pick what goes in.

04

Sign with the program

The license is between your lab and the program. Records pass directly between you, and the money follows your signed terms.

Public facts

What public numbers say about research data

These public facts describe other companies, never our own deals. We know of no biotech lab with a public, priced deal for its run records.

$110M+

Research publisher Wiley's total money from AI deals has passed this mark. Those deals cover published work, which differs from your lab notes. But they show AI firms pay for research content.

Wiley 8-K via SEC EDGAR, Jun 2026
$75M+

Academic publisher Taylor & Francis reported more than this from 2024 data access deals. Much of it came from AI companies. Its owner has called this kind of income non-recurring.

Informa 2024 results, Mar 2025, Informa 2025 results, Mar 2026

Timing

Why should a lab look at this now?

Nobody can honestly give you a date. These are better reasons to check.

Failed runs are unusual to find

Few people publish what went wrong. Programs tend to pay up for what no one else can supply.

Publishing changes the picture

Some programs want only unpublished data. Once a project comes out in a journal, that part of your archive may no longer qualify.

Labs close and archives scatter

When a project ends or a lab winds down, notebooks and instrument files can be lost. Closed labs can still qualify, if the records survive.

Prices may not hold

A rival holding similar records can cool demand for yours. No one can tell you how long current pricing holds.

Check your lab's fit

The quiz takes two minutes and asks nothing about your science. You will see an estimate and the programs that fit your lab.

See my lab's matches

Questions

Questions owners ask

Do failed experiments really have value?

Some programs ask for them by name. They want every run, with inputs and outcomes, including ones that failed. A model learns from what didn't work too. Please don't clean these out of your count.

We don't have 1,000 runs yet. Should we still check?

That number is what some programs ask for, and rules vary. Other programs look at office records instead, like email, chat and files. Those general programs often want about 50 desk staff, or about 20 or 30 full-time people in others.

We work with sponsors. Is any of it ours?

It depends on each agreement. Sponsored work often belongs to the sponsor, and some programs rule that out. Your own internal research may still be yours. Ask your lawyer to sort it project by project.

Does licensing hurt our patent position?

It could, depending on what the records reveal and when. We can't answer that for you. Talk to your patent lawyer before any record tied to an unfiled invention goes in. You can leave those projects out.

Can we include clinical trial data?

Some programs add value for clinical study reports with patient-level data. But sponsors, consent forms and HIPAA can all limit that. It must be de-identified, and you must own it. Ask your lawyer and your review board first.

Our lab has shut down. Is it too late?

Closed or closing companies can qualify, if the records still exist and someone has the right to sign. Check who holds the rights now. A license can't start until that is clear.

Does DillisData see our science?

No. We only see your quiz answers. Your notebooks, runs and files never reach us, and we couldn't share them if we wanted to.

Next step

Get an estimate for your lab's records

A quick quiz gives you a number we keep on the low side, from prices programs have published. Nobody is making an offer yet, and your notebooks stay put.

Free estimate. Two minutes. No uploads.