Mercor states a deal range of USD 50,000 to 875,000 for its research lab listing, checked 2026-11-09 and re-read 2026-11-17, and that range covers a whole archive, not a genome.
No public price per genome, sample or person is on record, and a genetic record also describes relatives who never gave consent. This page covers licensing only, as of 2026-11-17, and sits inside AI training data licensing. Not licensing is a valid outcome.
Figures on this page are claims by the stating party, checked 2026-11-09, and statute text was read on 2026-11-17. See how we source, date and check every figure.
What Counts as Genetic and Genomic Data in a Training Set
The laws a seller meets define genetic data differently: GINA counts information about a person's genetic tests and those of family members, the GDPR counts data on inherited or acquired genetic characteristics, and the DOJ rule sets a separate threshold for genomic data. The table compares the three definitions as of 2026-11-17.
| Instrument | What counts | Stated exclusion or threshold | Source |
|---|---|---|---|
| GINA, 42 U.S.C. 2000ff(4), (7) | Information about a person's genetic tests, family members' genetic tests and disease in family members; includes requests for genetic services and participation in clinical research with genetic services; a genetic test is an analysis of human DNA, RNA, chromosomes, proteins or metabolites that detects genotypes, mutations or chromosomal changes | Not sex or age; not protein or metabolite analysis that detects no genotypes, mutations or chromosomal changes | govinfo.gov |
| GDPR Art. 4(13), Art. 9(1) | Personal data on inherited or acquired genetic characteristics giving unique information on physiology or health, in particular from a biological sample | Special category; an Art. 9(2) exception is needed | eur-lex.europa.eu |
| 28 CFR 202.205(a) | Human genomic data; other human 'omic data | Bulk above 100 US persons (genomic) and 1,000 (other 'omic) in 12 months | ecfr.gov |
The record a holder keeps, whether raw sequence, genotype calls, test reports or family history, can fall under several definitions at once, while sex and age alone are not genetic information under GINA. The wider market for medical records, text and images is mapped in data types AI companies buy.
Who Asks for Genomic Records, and in Which Words
Mercor's Biotech & Life Sciences Lab listing is the one acquirer page on record that names genomic data: it lists consumer genomics companies and genomics labs among target industries and genomic data collected with research consent as nice to have. The table sets out the listing as Mercor states it, checked 2026-11-09 and re-read 2026-11-17.
| Listing (published) | Target industries and size | Requirements | Nice to have (genetic) | Not a fit | Deal range to the company, Mercor states | Source |
|---|---|---|---|---|---|---|
| Biotech & Life Sciences Lab (2026-10-17) | "Biotech and biopharma, CROs, genomics and diagnostics labs, consumer genomics companies, academic spin-outs"; 10 to 1,000 employees | Electronic lab notebooks or structured experiment records the organization owns; 1,000+ runs with inputs and measured outcomes, failed runs included; unpublished data | "Genomic data collected with research consent"; scoping question "Is any of it patient-level or genomic data, and under what consent?" | Data already published; data owned by a sponsor or partner; device or AI-software companies whose data is mainly software logs | USD 50,000 to 875,000, "Depends on volume and depth of records" | api.data.mercor.com listings/research-lab |
The listing asks the holder directly under what consent any genomic or patient-level data was collected, and it excludes data already published and data owned by a sponsor or partner. Terms, payouts and the program's record are in the Mercor data buyer profile.
No other program on record names genetic or genomic data, and a buyer that does not name it is not a buyer that accepts it. Buyers are named here only where their own pages ask for genomic records.
Genetic Databases Licensed or Sold on Record
Three transactions on record moved genetic databases, and a fourth, Tempus's license to Pathos AI, is the only one that states AI model training as its purpose, without stating genomic content.
The table lists them by date with values as the filing states them, and the full list sits in the tracker of AI data licensing deals.
| Date | Parties (seller; buyer) | Data | Use as stated | Value as stated | Source |
|---|---|---|---|---|---|
| 2022-01-18 | 23andMe Holding Co.; GSK plc | Genetic and health survey database | Drug target discovery; one-year extension of the exclusive period to July 2023 | USD 50 million one-time payment (23andMe 8-K exhibit) | sec.gov |
| 2023-10-30 | 23andMe Holding Co.; GSK plc | Genetic and phenotypic database | Drug target discovery and research; one-year non-exclusive data license including certain analysis services | USD 20 million upfront (23andMe 8-K exhibit) | sec.gov |
| 2025-04-17 | Tempus AI, Inc.; Pathos AI, Inc. | De-identified multimodal oncology dataset; genomic content not stated | Sole purpose: development and training of a foundation model | USD 200 million in data license fees over three years, including USD 50 million upfront (Tempus 10-K FY2025) | sec.gov |
| 2025-07-14 (closing) | 23andMe Holding Co. and debtor subsidiaries; TTAM Research Institute | Substantially all assets, including the genetic database | Bankruptcy asset sale, not an AI training license; sale order 2025-06-27 | USD 305.0 million in cash (23andMe 8-K) | sec.gov |
The two GSK plc agreements, filed by 23andMe on 2022-01-18 and 2023-10-30, concern drug target discovery, and AI training is not stated in either. The second is a one-year license that is non-exclusive, so it granted GSK plc no exclusive rights to the database.
The Tempus and Pathos license of 2025-04-17 covers a de-identified multimodal oncology dataset for the sole purpose of training a foundation model, at USD 200 million over three years including USD 50 million upfront (Tempus 10-K FY2025).
Whether the dataset holds genomic records is not stated; the filings are collected in Tempus AI clinical data licensing deals.
The fourth transaction is the 23andMe sale in In re 23andMe Holding Co., Bankr. E.D. Mo. 25-40976, in which substantially all assets including the genetic database went to TTAM Research Institute, a California nonprofit public benefit corporation, for USD 305.0 million in cash, with the sale order entered 2025-06-27 and the transaction consummated 2025-07-14 (status as of 2026-11-17).
It is an asset sale and not an AI training license, and the route is explained in data assets in insolvency.
Consent and Re-consent: Whether the Original Consent Reaches AI Training
The consent recorded when a sample was collected decides whether its data may be licensed for AI training, and a service or research consent that does not name third-party AI training usually needs new consent. Mercor's scoping question "under what consent?" and its preference for research consent show the same test from the acquirer's side.
In the EU, Article 9(2)(a) of the GDPR requires explicit consent for one or more specified purposes, and Article 7(3) lets the person withdraw consent at any time as easily as it was given (Regulation (EU) 2016/679, applies from 2018-05-25).
Article 6(4)(c) weighs the nature of the data, in particular special categories, when a new purpose is tested for compatibility with the original one.
In the US, FTC staff wrote on 2024-02-13 that it may be unfair or deceptive for a company to adopt more permissive data practices retroactively, a staff post and not a Commission rule.
The Oregon Attorney General's guidance of 2024-12-24 calls for explicit consent for sensitive data before training and rejects retroactive or passive notice changes; the contract side is covered in customer consent before licensing.
A sample's data also describes relatives who never consented, so the consent of the tested person does not cover them. That is a risk to weigh and not a legal conclusion.
Genetic Privacy Rules That Apply Before a License
Four sets of rules reach genetic records before any license, and this section is information, not legal advice.
GINA: Who the Federal Disclosure Bar Binds
GINA's disclosure bar binds employers, employment agencies, labor organizations and joint labor-management committees, which may not disclose an employee's or member's genetic information except in six listed cases (42 U.S.C. 2000ff-5(b), from Pub. L. 110-233, title II, s.206, 2008-05-21).
The six cases are disclosure to the employee or member on written request, to an occupational or other health researcher under 45 CFR part 46, under court order, to officials investigating compliance, for Family and Medical Leave Act certification, and to a public health agency for a contagious disease.
None of the six covers a commercial license, so an employer holding staff genetic records cannot license them. The bar names no testing company or laboratory, and this page states nothing on the Act's other titles.
State Rules That Name Genetic Data: Washington, Oregon and Connecticut
Washington, Oregon and Connecticut each name genetic data in their privacy statutes and require the person's authorization or consent before it is sold. Washington's rule sits in Washington's health data law, and the table gives all three as of 2026-11-17.
| State (statute, in force) | Where genetic data sits | Rule before a sale | Source |
|---|---|---|---|
| Connecticut (CTDPA, Conn. Gen. Stat. 42-515 to 42-525; amendments 2026-07-01 and 2026-10-01) | Sensitive data (42-515); from 2026-10-01 genetic data is not "publicly available" unless the consumer made it public | No sale of sensitive data without the consumer's consent (42-520(a)(1)(H)), from 2026-07-01 | cga.ct.gov |
| Oregon (OCPA, ORS 646A.570 to 646A.589; 2024-07-01) | Sensitive data (646A.570(18)) | No processing of sensitive data without consent (646A.578(2)(b)); AG guidance of 2024-12-24: explicit consent before training, no retroactive notice changes | doj.state.or.us |
| Washington (My Health My Data Act, RCW 19.373; 2024-03-31) | Consumer health data (19.373.010(8)(b)(x)) | Valid authorization signed by the consumer naming the buyer (19.373.070), valid one year, kept six years by seller and purchaser | app.leg.wa.gov |
Illinois and California rules on genetic information are not stated on this page, and other states are outside it until their statutes are verified.
Genetic Data Under the GDPR: Articles 4(13), 9(1) and 9(4)
Genetic data is special-category data under GDPR Article 9(1), so a license needs an Article 9(2) exception in addition to an Article 6 basis, and Article 9(4) lets member states add conditions, including limitations, for genetic data. The exceptions are set out in special category data in AI training.
The proposed Article 9(2)(k) in the Digital Omnibus, COM(2025) 837, is pending and is not law.
Bulk Genomic Data and Buyers Linked to Countries of Concern
Human genomic data about more than 100 US persons in any 12 months is bulk data under 28 CFR 202.205, the lowest threshold in the Department of Justice rule (eCFR as of 2026-11-05).
Licensing it to a buyer in or linked to China, Cuba, Iran, North Korea, Russia or Venezuela is prohibited since 2025-04-08, even when the data is anonymized or de-identified.
Other foreign buyers need the onward-transfer clause under 202.302. Thresholds and foreign buyers are covered in the DOJ bulk sensitive data rule.
Preparing Genetic Records: Consent Files, Exclusions and De-identification Limits
A genetic dataset is ready to offer only when each record can be matched to its consent, records owned by a sponsor or partner are out, and the seller knows which rules still apply after identifiers are removed. The checks run in this order:
- A consent record for every sample, showing the purpose it names.
- Sponsor- or partner-owned data removed, as Mercor's not-a-fit list requires.
- Published data set aside, because Mercor does not take it.
- Staff genetic records held as an employer excluded, under the GINA bar.
- A count of US persons per category against the 202.205 thresholds.
De-identification does not take data out of the DOJ rule, which covers anonymized and de-identified data. No verified de-identification standard for genomic data is on record, so this page states none.
Removing names and contact fields is only part of the work, and the methods are set out in PII removal and de-identification.
Licensing Genetic Data: Benefits, Risks, Downsides and Alternatives, Including Not Licensing
Licensing genetic data can earn from archives a lab or company already holds, but consent, relatives' interests and genetic privacy law weigh more here than for any other company data, and a model trained on it cannot be untrained.
Benefits
- A named demand exists: Mercor lists consumer genomics companies and research-consented genomic data.
- A non-exclusive license, the form of the 2023-10-30 GSK plc agreement, grants the buyer no exclusive rights.
- Filings show that genetic databases carry value, from the 2023 GSK plc upfront fee to the 23andMe asset sale price.
Risks
- A consent that does not name AI training, and FTC staff and Oregon warnings on retroactive changes.
- State authorization and consent rules in Washington, Oregon and Connecticut.
- The DOJ rule at a threshold of more than 100 US persons.
- Insolvency: the 23andMe genetic database passed with substantially all assets in a court-supervised sale.
Downsides
- Mercor's not-a-fit list excludes published data and sponsor or partner data.
- No public price benchmark per genome or sample is on record.
- A trained model keeps what it learned, so deletion later does not reverse it.
Alternatives
- License lab notebooks and run records without genomic data, which Mercor lists as included.
- A non-exclusive, fixed-term license, the form of the 2023 GSK plc agreement.
- A research collaboration under research consent, the GINA exception for researchers under 45 CFR part 46.
- Not licensing at all, which remains a valid outcome.
The consent file comes before any price.
Neighboring Health Data Types and the Seller Steps for Genetic Records
This page covers genetic and genomic records; patient charts, imaging, research papers, faces and voices each have their own page, and HIPAA's de-identification rules are covered with clinical records.
- Medical and clinical data: patient records and HIPAA.
- Medical imaging and pathology data: imaging studies and slides.
- Scientific data and papers: research records and papers.
- Universities and research groups licensing data: research holders and consent.
- Biometric data in AI training: faces and voices.
- HIPAA and AI training data: HIPAA de-identification.
- European Health Data Space: the EU permit route.
- Tempus AI clinical data licensing deals: the Tempus filings.
- US AI training data law: federal and state rules.
The seller steps for genetic records follow the general process:
- Read the guide on how to sell or license company data, the complete seller guide.
- Check the list of company data that cannot be licensed across sectors.
- Review AI data licensing agreements for every license clause.
- Plan for after the sale: deletion and reuse limits.
Frequently Asked Questions
Does GINA stop a company from licensing genetic data?
Not by itself: GINA's Title II bar binds employers and the other listed bodies, not every holder. State and EU rules may still apply to a testing company or laboratory.
Is de-identified genomic data outside the DOJ bulk data rule?
No, the rule covers anonymized and de-identified data. The genomic threshold is more than 100 US persons in any 12 months (28 CFR 202.205, eCFR as of 2026-11-05).
Can genetic data collected for ancestry or health reports be reused for AI training?
Only where consent or law allows the new purpose. Retroactive notice changes draw warnings from FTC staff (2024-02-13) and the Oregon Attorney General (2024-12-24).
How much do AI companies pay for genetic data?
No price per genome or sample is on record. Mercor states a deal range of USD 50,000 to 875,000 for its lab listing (checked 2026-11-09), and the GSK plc filings state USD 50 million and USD 20 million; all are claims about whole deals. Methods are in the guide to what company data is worth.