Biotechs, CROs and contract labs hold assay data that AI developers ask for: ELN and LIMS records linked to results, failed runs included. A CRO usually runs studies for sponsors, and in Origin's view most services agreements give the sponsor control of those results; the agreement decides. The lab's own methods and internal research may be its own. No closed price for an assay archive has been published.
- Results linked to the protocol, the instrument file and the failed attempts are what make assay data useful.
- At a CRO, the services agreement decides who controls each study: in Origin's view usually the sponsor, not the lab.
- GLP and Part 11 records must stay intact and retrievable; a licensed copy never replaces the archive.
- Human-subject data needs a check of consent scope, HIPAA and the Common Rule before anything else.
- For a closing biotech, the board signs while the company exists; an assignee or trustee signs after.
What assay data do AI developers ask for?
Labs building models for biology and chemistry ask for experiments with their context: what was tested, how, on which instrument, what came out and whether it worked. OpenAI's science data role describes work that "could range from acquiring existing datasets to shaping research collaborations" and says the current focus is "particularly on biology and chemistry" (job post). Anthropic is hiring for a Life Sciences team aimed at "accelerating progress in the life sciences, from early discovery through translation" (job post).
Failed experiments matter because models learn little from a record of only the runs that worked. The existing guide to licensing R&D lab notebooks and experiment records covers that general case; this page covers what is specific to CROs, contract labs and closing biotechs.
| Record | What it shows | Why it matters |
|---|---|---|
| ELN experiment entries | Hypothesis, protocol version, observations, conclusion | The reasoning behind each run |
| LIMS results | Sample, method, value, units, pass or fail | Structured outcomes that can be joined to entries |
| Instrument files and images | Plate reads, chromatograms, blots, histology | The raw evidence behind each number |
| Failed and repeated runs | What went wrong and what changed | Negative results that rarely reach a paper |
| Study reports | Protocol, methods, statistics, conclusions | A finished, reviewed account of a whole study |
Who owns study data at a CRO or contract lab?
Holding the data is not the same as controlling it. In Origin's view of typical practice, a services agreement or work order gives the sponsor the results of studies it pays for, while the lab keeps its own methods, SOPs, validation work and internal research. Your agreements decide the real position, study by study.
The regulations point the same way. Under the IND rules, "A sponsor may transfer responsibility for any or all of the obligations set forth in this part to a contract research organization" and "Any such transfer shall be described in writing" (21 CFR 312.52). Under GLP, a sponsor is "A person who initiates and supports, by provision of financial or other resources, a nonclinical laboratory study" (21 CFR 58.3), and if a testing facility closes, its study records "shall be transferred to the archives of the sponsor of the study" (21 CFR 58.195(h)).
This page is general information, not legal advice: it quotes each rule with its source, and your counsel decides how it applies. A lawyer should read your services agreements, material transfer and collaboration agreements, consent forms and any estate papers before anything is shared.
| Holder | Usual position on the data | Can it sign alone? |
|---|---|---|
| Biotech running its own research | Owns its own experiments | Yes, through its officers or board, subject to partner agreements |
| CRO or contract lab, sponsor studies | Sponsor controls results under the services agreement | Usually not; needs the sponsor's written consent |
| CRO or contract lab, own work | Methods, SOPs, validation and internal R&D | Often yes, if no client data is inside |
| Academic or partner collaboration | Shared under a collaboration or material transfer agreement | Only as the agreement allows |
| Closed biotech | The company, then its assignee or trustee | The board, assignee or trustee, depending on the stage |
What is assay data worth? The evidence, graded
No closed price for an archive of assay data has been published. The figures below are labour rates, prices for generating new data and one whole-company sale. They are ceilings or reference points, not offers.
| Evidence | What it says | Grade |
|---|---|---|
| OpenAI science data role | Acquiring existing datasets, focused on biology and chemistry (OpenAI) | Company claim (job post); no price |
| Anthropic life sciences | Life Sciences research engineering team (Anthropic) | Company claim (job post); no price |
| Mercor drug discovery experts | "$70-$100 per hour" for ADME/Tox and pharmacology experts working for a leading AI lab (Mercor) | Company claim (listing); a labour rate, not a data price |
| Handshake Vials benchmark | Top models "sub-30% on Vials, under 60% even with tools" on life-sciences image tasks (podcast summary) | Secondary summary of a company claim |
| Ginkgo Bioworks, Q3 2026 | Says it is "generating biological and chemistry data for several frontier AI labs" (call summary) | Secondary summary of an earnings call; a counter-signal: labs order new data |
| Ginkgo Datapoints ADME-One | New ADME profiling at "$199/compound" (Ginkgo Datapoints) | Company claim (price list); in Origin's view, a ceiling for archived ADME data on the same compounds |
| 23andMe bankruptcy sale | Court approval to sell its "genetic data trove" for $305 million (Bloomberg Law) | Independent press on a court record; genomics plus a whole company, not assay data |
Read these as context, not as a price for your archive. What moves value is years of linked experiments with metadata, including the failures, that the buyer could not cheaply generate again.
Which rules limit licensing lab and study data?
GLP records stay in the archive. Under 21 CFR Part 58, "All raw data, documentation, protocols, final reports, and specimens" from a nonclinical laboratory study "shall be retained", in archives "for orderly storage and expedient retrieval" (21 CFR 58.190), for the periods in 58.195. A licensed copy is a copy; the original archive and its index stay where the rule puts them.
Electronic records under FDA rules fall under Part 11, which "applies to records in electronic form that are created, modified, maintained, archived, retrieved, or transmitted, under any records requirements set forth in agency regulations" (21 CFR 11.1). Exporting a copy for a licensee must not break the audit trail or the signed record in the source system.
Human-subject data needs its own review. If your lab is a HIPAA covered entity or business associate, "a covered entity or business associate may not sell protected health information" except with the individual's authorization (45 CFR 164.502(a)(5)(ii)). De-identified information, by Expert Determination or by removing the Safe Harbor identifiers, is health information that does "not identify an individual" (45 CFR 164.514). The Common Rule (45 CFR 46) "applies to all research involving human subjects conducted, supported, or otherwise subject to regulation by any Federal department or agency" that adopts it (45 CFR 46.101), and it covers "identifiable private information" (45 CFR 46.102). Clinical investigations regulated by FDA, including privately funded trials a CRO runs for a sponsor, fall under FDA's own human-subject rules whether or not the Common Rule applies: 21 CFR Part 50 on informed consent "applies to all clinical investigations regulated by the Food and Drug Administration" under sections 505(i) and 520(g) of the act (21 CFR 50.1), and Part 56 sets the standards for an IRB "that reviews clinical investigations regulated by the Food and Drug Administration" (21 CFR 56.101). Where none of these applies, the consent forms behind samples and data still set what they may be used for.
Material transfer agreements and collaboration agreements often limit how data generated with a partner's materials may be used or shared, and some give the partner rights in the results. List every one that touches a project before describing it.
This page is general information, not legal advice: it quotes each rule with its source, and your counsel decides how it applies. A lawyer should read your services agreements, material transfer and collaboration agreements, consent forms and any estate papers before anything is shared.
Who signs when a biotech closes?
While the company exists, its board decides, usually on the officers' recommendation. In an assignment for the benefit of creditors, the assignee holds the assets and decides; see selling company data in an ABC. In Chapter 7 the trustee decides, and a sale outside the ordinary course needs the court: "The trustee, after notice and a hearing, may use, sell, or lease, other than in the ordinary course of business, property of the estate" (11 U.S.C. 363(b)(1)). See selling a debtor's data in a 363 sale.
Before anything closes, check the GLP rule: if a testing facility goes out of business, its study records "shall be transferred to the archives of the sponsor of the study" and FDA "shall be notified in writing of such a transfer" (21 CFR 58.195(h)). Sponsor studies leave with the sponsor; the company's own research is what an assignee or trustee may be able to license.
Origin pays no fee or other payment to a trustee, debtor in possession, assignee for the benefit of creditors, receiver, examiner or estate counsel or other professional, for an estate they serve. Their duties run to the estate and its creditors. The estate pays Origin nothing; a buyer may pay Origin when a deal closes, and any payment Origin receives from a buyer is disclosed to the estate, so the estate can disclose it to the court.
Records checklist for a biotech, CRO or contract lab
The ELN and LIMS export checklist (CSV) has the system-level rows; these are the questions to answer first.
| Item | What to write down |
|---|---|
| Systems | ELN, LIMS, instrument data stores, study archive; years in each |
| Volume | Experiments and studies per year, including failed and repeated runs |
| Links | Whether ELN entries, LIMS results and instrument files share sample or study IDs |
| Ownership | Which studies were run for sponsors, which under collaboration or material transfer agreements, which are your own |
| Regulated records | GLP studies and their archive locations; Part 11 systems |
| Human-subject data | Any patient or donor data, consent forms, IRB approvals, HIPAA status |
| Exclusions | Sponsor studies without consent, partner-restricted projects, identifiable data |
What a one-page datasheet looks like (fictional example)
Invented to show the level of detail; it describes no real company.
| Field | Fictional example |
|---|---|
| Company | Preclinical biotech, small-molecule kinase program, 45 staff, closing |
| Years | ELN 2018 to 2026; LIMS 2020 to 2026 |
| Volume | About 31,000 experiment entries; 1.2 million assay results; 40% marked failed or repeated |
| Rights | Own research only; two partner-funded projects excluded under their collaboration agreements |
| Exclusions | No human-subject data; staff names coded |
| Signer | Board before the assignment; assignee after |
What happens next: three steps
A first conversation needs no files. You can stop after any step.
- Describe your records: answer six questions, with no files and no system access. Origin then checks which studies are your own, which belong to sponsors or partners, and what consent or agreement limits apply.
- Datasheet and sample: Origin turns your answers into a one-page datasheet like the example above, and you approve a small de-identified sample before anything leaves the company.
- Buyers and license: Origin shows the datasheet to the AI developers and data programs that have asked for assay and study data, with their published terms. You approve scope, use, price and delivery, and sign the license yourself.
- Origin does not charge data owners; a buyer may pay Origin when a deal closes (how Origin is paid).
Resources for lab managers, QA and study directors
Guides for the people who keep these records:
- IC50: how to calculate it, with a 96-well worked example
- 21 CFR Part 11: requirements and a compliance checklist
- 21 CFR Part 58 (GLP): final reports, archives and retention
- ELN vs LIMS: differences and what each exports
- Licensing R&D lab notebooks and experiment records
- How to de-identify company data
Hold years of lab notebooks and assay results? Describe them in six questions, no files needed, and see what your data is worth.
Origin Data Partners works with companies that own data and helps get it bought. We are independent of the services on this page, and a buyer may pay us when a deal closes (how Origin is paid). See what your data is worth.
Common questions
Can a CRO license the study data it generated for sponsors?
Usually not on its own. In Origin's view of typical services agreements, the sponsor controls the results of the studies it pays for, so the CRO needs the sponsor's written consent. Its own methods, SOPs, validation work and internal research may be its own to license.
Are failed experiments and negative results worth anything to AI developers?
They can be, when they are linked to the protocol, the instrument data and what changed on the next attempt. Few negative results are published, so a lab's record of what did not work is hard to find elsewhere.
What happens to GLP study records when a testing facility goes out of business?
Under 21 CFR 58.195(h), the raw data, documentation and other required material are transferred to the archives of the sponsor of each study, and FDA is notified in writing. Plan this before a wind-down, not after.
Who can sign a data license for a biotech that is shutting down?
The board while the company exists. After an assignment for the benefit of creditors, the assignee; in Chapter 7, the trustee, with court approval for a sale outside the ordinary course. Origin pays none of them a fee for an estate they serve.
Does a material transfer agreement limit what we can do with our results?
Often. Many material transfer and collaboration agreements limit use of results generated with the partner's materials or give the partner rights in them. Read each one that touches a project before you describe that project to anyone.
Does the Common Rule apply to a private biotech's human-subject data?
It applies to research conducted, supported or regulated by a federal department or agency that has adopted it, so federally funded work is covered. Clinical investigations regulated by FDA fall under 21 CFR Parts 50 and 56 instead, even when privately funded. Either way, the consent forms and IRB approvals behind the data set what it may be used for.
What is assay data worth if no price has been published?
No closed price for an assay archive has been published. Reference points are labour rates for drug-discovery experts, such as Mercor's listed $70 to $100 per hour, and the cost of generating new data, such as Ginkgo's listed $199 per compound for ADME profiling. Neither is an offer.
What does Origin charge a lab or biotech?
Nothing. Origin does not charge data owners; a buyer may pay Origin when a deal closes. You approve every step and sign the license yourself.
Sources and further reading
- 21 CFR 312.52: transfer of obligations to a contract research organization
- 21 CFR 58.3: definitions (nonclinical laboratory study, sponsor, raw data)
- 21 CFR 58.190: storage and retrieval of records and data
- 21 CFR 58.195: retention of records
- 21 CFR 11.1: scope of Part 11 (electronic records and signatures)
- 45 CFR 164.502: uses and disclosures of protected health information (sale of PHI)
- 45 CFR 164.514: de-identification of protected health information
- 45 CFR 46.101: to what the Common Rule applies
- 45 CFR 46.102: Common Rule definitions (human subject, identifiable private information)
- 21 CFR 50.1: scope of FDA's informed consent rules (eCFR)
- 21 CFR 56.101: scope of FDA's institutional review board rules (eCFR)
- 11 U.S.C. 363: use, sale or lease of property of the estate
- OpenAI careers: Data Program Manager, Science (job post)
- Anthropic careers: Research Engineer, Life Sciences (job post)
- Mercor jobs: Drug Discovery Specialist, Pharmacology & ADME/Tox (listing)
- BigGo Finance: podcast summary on Handshake AI benchmarks (Vials, Atlas)
- TradingKey: Ginkgo Bioworks Q3 2026 earnings call summary
- Ginkgo Datapoints on LinkedIn: ADME-One launch post
- Bloomberg Law: 23andMe wins approval of $305 million genetic data asset sale (June 29, 2025)