- Who owns trial data is set by each clinical trial agreement. One published US template gives the sponsor the study data "in accordance with the signed informed consent and authorization form", while source records stay with the site.
- FDA rules (21 CFR parts 50, 56, 312, 812 and 11) govern consent, IRB review, record retention and electronic records. We found no provision in them that grants or bars a license of trial data.
- Results posted on ClinicalTrials.gov are summary data. The consent statement FDA requires tells participants the ClinicalTrials.gov website "will not include information that can identify you".
- No sponsor has published a price for licensing clinical trial data to AI companies. Existing sharing platforms give researchers access under data use agreements, not for a fee per record.
Who may license clinical trial data, and under which rule?
Three layers decide it: the contracts between sponsor, site, CRO and investigator; the consent and HIPAA authorization each participant signed; and the regulations on records and privacy. The table takes each holder in turn and quotes the governing words. Consent runs through every row, because the agreements and the rules both point back to what participants agreed to.
| Holder and rule or law | Operative words (quoted) | Plain meaning for a license | Source and date read |
|---|---|---|---|
| Sponsor: the clinical trial agreement (Accelerated Clinical Trial Agreement template, section 4) | "Sponsor shall own and have the right to use the Data in accordance with the signed informed consent and authorization form, applicable laws, and the terms of this Agreement." | Under this template the sponsor owns the study data, but its use is bounded by the consent form, the law and the contract. Other agreements differ | ara4us.org template updated September 2025, read 20 November 2026 |
| Site or institution: the same clinical trial agreement | Medical records and source documents "shall remain the sole and exclusive property of the Institution or medical provider"; the site keeps rights for "noncommercial research purposes" | Under this template a site may not license the study data commercially; its patient records are its own and are governed by HIPAA | ara4us.org, read 20 November 2026 |
| Site that is a HIPAA covered entity: 45 CFR 164.514 and 164.508 | A research authorization must warn of "redisclosure by the recipient and no longer be protected by this subpart"; HHS counts "Clinical trial record numbers" as identifiers | Sites may license patient data only after de-identification or with authorization. A sponsor that is not a covered entity is usually outside HIPAA once it holds the data | eCFR and HHS, read 20 November 2026 |
| Investigator and IRB in FDA-regulated trials: 21 CFR 50 and 21 CFR 56 | Except as 50.22 to 50.24 provide, no investigator may involve a subject without "the legally effective informed consent of the subject or the subject's legally authorized representative"; trials need IRB approval and "continuing review" | The exceptions include an IRB waiver or alteration of consent for minimal-risk investigations (50.22, added December 2023). These rules govern consent and review. We found no provision in them that grants or bars a later license; the consent form's own words govern future use | eCFR, read 20 November 2026 |
| Drug and device trials: 21 CFR 312 (IND) and 21 CFR 812 (IDE) | Drug investigators keep records "for a period of 2 years following the date a marketing application is approved" or, if none is filed or approved, "until 2 years after the investigation is discontinued and FDA is notified" | Sponsors have parallel duties (312.57). Device records are kept 2 years after the later of completion or the date they stop supporting an FDA submission (812.140(d)). Licensing a copy does not change these duties; the originals stay available for inspection | eCFR, read 20 November 2026 |
| Electronic trial records: 21 CFR 11 | Part 11 "applies to records in electronic form that are created, modified, maintained, archived, retrieved, or transmitted, under any records requirements" | Part 11 governs the trustworthiness of regulated records. It says nothing about licensing; an export for a buyer must not alter the source system | eCFR, read 20 November 2026 |
| Responsible party for an applicable trial: 42 U.S.C. 282(j) and 42 CFR 11 | Results include "The number of human subjects that started and completed the clinical trial, by arm." | ClinicalTrials.gov results are summary counts and measures by arm. Posting them licenses nothing and releases no participant-level records | eCFR, read 20 November 2026 |
| Federally supported research: Common Rule, 45 CFR 46 | Secondary research use is exempt under 46.104(d)(4) only if a listed route applies, such as information recorded so identity "cannot readily be ascertained directly or through identifiers linked to the subjects, the investigator does not contact the subjects, and the investigator will not re-identify subjects" | Another route covers uses regulated by HIPAA as research or health care operations. Without an exemption, reuse of identifiable trial data for new research needs consent, broad consent or an IRB waiver | eCFR, read 20 November 2026 |
| Any US holder with a foreign licensee: 28 CFR 202 | Exempts transactions only "to the extent" they are "Ordinarily incident to and part of clinical investigations regulated by the U.S. Food and Drug Administration" (202.511); 202.510 exempts regulatory approval data where "necessary to obtain or maintain regulatory authorization" | Neither text mentions a separate commercial license for AI training, and we found no Justice Department guidance applying them to one. Data sent to one foreign person in a deal or over 12 months is bulk, even when de-identified, if it includes human genomic data on more than 100 US persons, other 'omic data on more than 1,000 or health data on more than 10,000. Licensing bulk data to a country of concern or covered person is prohibited | eCFR, read 20 November 2026 |
| Sponsor with EU participants: GDPR Articles 6, 9 and 44 and EU Regulation 536/2014, Article 28(2) | A sponsor may ask participants "to consent to the use of his or her data outside the protocol of the clinical trial exclusively for scientific purposes"; the EDPB says that consent "is not the same consent referred to in the GDPR as one of the legal basis" | Secondary use of EU trial data needs an Article 6 basis and an Article 9(2) condition, any member state conditions under Article 9(4), and a Chapter V transfer mechanism for licensees outside the EEA. Coded trial data is pseudonymized, not anonymous, under Recital 26. Whether commercial AI training is a scientific purpose is unsettled | EUR-Lex and EDPB, read 20 November 2026 |
Sources: the ACTA template, 45 CFR 164.508, HHS de-identification guidance, 21 CFR 50.20, 21 CFR 56.103, 21 CFR 312.62, 312.57, 50.22, 21 CFR 812.140, 21 CFR 11.1, 42 CFR 11.48, 45 CFR 46.104, 28 CFR part 202 and Regulation 536/2014 and EDPB Opinion 3/2019, read on 20 November 2026. This page is general information, not legal advice.
Who owns trial data: the sponsor, the site or the investigator?
Usually the sponsor owns the study data, the site owns the medical records, and the investigator owns neither and works under the institution's rights, including its right to publish. That is the split in the Accelerated Clinical Trial Agreement, a standard template for sponsor-initiated multisite trials published by the Accelerated Research Agreements initiative. It defines "Data" as what the institution generates "in the performance of the Study and required to be delivered" under the protocol, and excludes patient medical records, research notebooks and source documents.
The template attaches a condition to the sponsor's ownership that matters for AI: the sponsor owns and may use the data "in accordance with the signed informed consent and authorization form, applicable laws, and the terms of this Agreement". Ownership does not override consent. The institution keeps the right to use the data "for publication, IRB, regulatory, legal, and for its own internal educational, patient care, and noncommercial research purposes".
Many sponsors use their own forms, and terms differ by sponsor, institution and country. Read the data clause, the definitions it relies on, any license back to the site, the confidentiality clause and the survival clause. An investigator-initiated trial is different again: there the institution is often the sponsor, and its own policies and grant terms decide.
Can we sell our clinical trial data to AI companies?
A sponsor may license trial data to an AI company only if four things line up: it owns or controls the data under its site agreements, the consent forms cover the use, the data is de-identified to the standard the license requires, and no partner or co-development agreement gives someone else exclusive rights. A license grants rights to a de-identified copy. It does not transfer the underlying records, which must be kept for FDA.
Sponsors already share trial data with outside researchers, and the terms show what buyers expect. Vivli's data use agreement says the "Recipient agrees that it will restrict its use of any Data Set to the Analysis described in the Research Plan", forbids intentional attempts to identify individuals who are subjects of the data, and keeps "any pre-existing ownership rights" with the data contributor. The YODA Project says "A signed Data Use Agreement (DUA) is required for data access" and that each data partner has its own DUA. These are research access routes under approved plans, not commercial licenses, but a commercial license will be read against the same expectations.
Two practical limits apply to most archives. Trials with EU sites bring the GDPR and Article 28 of the EU clinical trials regulation. Trials run before current consent templates may have forms that limit data use to the study itself, which narrows what can be licensed without new consent or an IRB decision.
Does trial participants' informed consent cover AI training?
Only if its words do. Read three documents for each trial: the informed consent form, the HIPAA authorization (often combined with it), and any optional consent for future research or sample storage. The federal research rules give the forms a vocabulary to look for.
For research the Common Rule covers, 45 CFR 46.116(b)(9) says a consent form for research that collects identifiable information must state either that de-identified data "could be used for future research studies or distributed to another investigator for future research studies without additional informed consent", or that it will not. Where appropriate, it should also say whether specimens "may be used for commercial profit" and whether participants share in that profit. A form with the first statement supports research reuse. Whether "future research" covers commercial AI training is not settled by the rule, and IRBs and counsel answer it trial by trial.
HIPAA research authorizations may run to "end of the research study" or "none", including for a "research database or research repository". The authorization must also warn that information disclosed to the sponsor may be redisclosed "and no longer be protected" by HIPAA. That warning describes the law; it does not give the sponsor permission for uses the consent form did not describe.
Can a contract research organization license its lab data for AI training, or does the sponsor own it?
For study results, the sponsor almost always controls them. A CRO runs work on the sponsor's behalf under a master services agreement and work orders that typically assign study data and results to the sponsor and treat them as the sponsor's confidential information. A CRO may license study data only if its agreement with that sponsor gives it the right, which is uncommon. The biotech, CRO and lab data guide covers lab and assay data in more depth.
What a CRO may own is the data about its own operations: how long sites took to activate, query rates, monitoring findings, protocol deviation patterns, validated SOPs and its own methods. Even there, check two things. Confidentiality clauses often reach information derived from a sponsor's study, and operational metrics tied to a named study or site can reveal the sponsor's confidential information. Stripping study identifiers and aggregating across sponsors, with each agreement checked, is the usual starting point.
Can a sponsor sell clinical trial enrollment data?
Not identifiable enrollment data without each person's authorization. Screening logs, enrollment lists and contact details are held by sites, and at a site that is a covered entity they are protected health information. HIPAA says "a covered entity or business associate may not sell protected health information" except with an authorization that states the holder is paid. Sponsors usually receive coded data, and a coded participant number is itself an identifier: HHS notes that "Clinical trial record numbers" fall under the safe harbor catch-all.
What a sponsor may license, subject to its agreements, is de-identified or aggregate enrollment information: screen-failure reasons, enrollment rates by site type, eligibility criteria that excluded most patients and how long recruitment took. Information like that is the kind a trial-design or recruitment model would use, though no buyer has published what it pays for it.
How do you license clinical trial data for AI training?
We help you license your data. Origin Data Partners is an independent data partner that helps business owners value and license their company records: the company pays Origin nothing, and a buyer may pay Origin when a deal completes. For a sponsor, the work usually runs in this order, with privacy counsel and, where needed, an IRB deciding the questions marked below.
- List the trials: phase, indication, years, countries, number of participants, and whether each was an applicable clinical trial under 42 CFR 11.
- Pull the site agreement template used for each trial and confirm the sponsor owns the study data and that no partner holds exclusive rights.
- Collect every consent form version and HIPAA authorization, and classify each trial: future research covered, silent, or excluded. Counsel or the IRB decides the silent ones.
- Separate EU, UK and other non-US participants and check the legal basis for their data under the GDPR and Article 28 of Regulation 536/2014.
- De-identify the participant-level data to the standard the license needs, usually expert determination, and keep the source records and audit trails untouched (see the Part 11 checklist).
- Write the license: permitted uses, no re-identification, no onward transfer to countries of concern, security, term, and what happens if a regulator asks about the data.
- Check vendors too: under 28 CFR 202.401 a vendor, employment or investment deal giving a country of concern or covered person access to bulk health data is a restricted transaction with security duties, and for bulk 'omic data it is prohibited.
When you have a list of trials and their consent status, check what your trial data may be worth.
How much is clinical trial data worth to AI companies?
No sponsor or buyer has published a price for licensing clinical trial data to AI companies, and we found no public per-participant or per-trial figure. Platforms such as Vivli and the YODA Project give researchers access under data use agreements tied to approved research plans; their published terms are about use limits, not price.
What moves value in a private negotiation is easier to list than the number. Participant-level data with protocol-defined endpoints, adverse events and lab values over time is scarcer than summary results, which are public for applicable trials. Failed and terminated trials matter because they show what did not work. Consent that clearly covers future research, clean de-identification and no EU or partner constraints shorten the buyer's review. Data from therapy areas with few public trials is harder to find elsewhere.
How we checked this
We read every source on this page on 20 November 2026. Regulations come from the eCFR (21 CFR parts 11, 50, 56, 312 and 812; 42 CFR part 11; 45 CFR parts 46 and 164; 28 CFR part 202), HHS guidance on de-identification and EUR-Lex for the GDPR and Regulation 536/2014. Contract terms come from the published Accelerated Clinical Trial Agreement template and the Vivli and YODA Project data use pages. We checked for 2025 and 2026 changes and found an HHS regulatory agenda entry planning Common Rule revisions, still at the proposed rule stage with no text we could find. The Justice Department rule has been in force since 8 April 2025; its compliance-program and audit duties for restricted transactions apply from 6 October 2025.
What we could not verify: the data clauses in any individual sponsor's site agreements or CRO contracts, which are not public; how IRBs have ruled on AI training under "future research" language, since those decisions are rarely published; and any price paid for trial data, because none has been published. This page does not cover 42 CFR part 2 for substance use treatment records, pediatric research rules, the European Health Data Space Regulation or information blocking rules.
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
Are results posted on ClinicalTrials.gov participant-level data?
No. The results a responsible party must post under 42 CFR 11.48 are summary information, such as participant flow, baseline characteristics and outcome measures by arm. The consent statement FDA requires for applicable trials tells participants the website "will not include information that can identify you. At most, the Web site will include a summary of the results." Posting them does not license any participant-level records.
Does Part 11 apply to a trial dataset exported for licensing?
Part 11 applies to electronic records kept under FDA record requirements or submitted to FDA. A copy made only for a licensee is generally not a record FDA requires, but the source system is. The export must not change, delete or overwrite the regulated records or their audit trails, and the originals must remain available for retention and inspection for the periods 21 CFR 312.62 and 812.140 set.
Do EU trial participants need to consent separately to secondary use?
Often. Article 28(2) of Regulation 536/2014 lets a sponsor ask participants, when they consent to the trial, to consent to use of their data "outside the protocol of the clinical trial exclusively for scientific purposes". The EDPB says that consent is not itself the GDPR legal basis, so the sponsor also needs an Article 6 basis and an Article 9(2) condition. Whether commercial AI training counts as a scientific purpose is unsettled and depends on the facts and the member state.
Can an investigator license data from a trial they ran at their site?
Rarely on their own. Under common templates the sponsor owns the study data and the institution owns the medical records, while the investigator works under the institution's agreement. Even the institution's retained rights are limited to purposes such as publication and "noncommercial research". An investigator who sponsored the trial is in a different position, governed by the institution's policies, grant terms and consent forms.
Do Vivli and the YODA Project allow trial data to be used for AI?
Their agreements allow the use an approved research plan describes. Vivli's agreement limits recipients to "the Analysis described in the Research Plan" and forbids intentional attempts to identify participants. The YODA Project uses a data use agreement specific to each data partner. A machine learning analysis can be a research plan, but a commercial model built for resale would need terms beyond an approved research plan.
Is a clinical trial record number a HIPAA identifier?
Yes, under safe harbor. HHS guidance quotes the Privacy Rule preamble: "Clinical trial record numbers are included in the general category of" any other unique identifying number, characteristic or code. A dataset that keeps participant numbers linked to site records is not de-identified by safe harbor. A covered entity may keep a re-identification code only if it meets the conditions in 45 CFR 164.514(c).
Does a sponsor need IRB approval to license old trial data?
It depends on the data and who funds or regulates the use. Reusing identifiable data for new research under the Common Rule needs consent, broad consent, an exemption or a waiver. De-identified data licensed outside federally supported research may need no IRB review under the federal rules, but the original consent terms still apply, and many sponsors ask an IRB anyway.
Sources and further reading
- Accelerated Research Agreements initiative: Accelerated Clinical Trial Agreement template (updated September 2025, PDF)
- Accelerated Research Agreements initiative (ARA4US)
- eCFR: 45 CFR 164.514, de-identification of protected health information
- eCFR: 45 CFR 164.502, including (a)(5)(ii), sale of protected health information
- eCFR: 45 CFR 164.508, uses and disclosures for which an authorization is required
- HHS Office for Civil Rights: guidance on HIPAA de-identification methods
- eCFR: 21 CFR 50.20, general requirements for informed consent
- eCFR: 21 CFR 50.22, exception from informed consent for minimal risk clinical investigations
- eCFR: 21 CFR 50.25, elements of informed consent
- eCFR: 21 CFR 56.103, circumstances in which IRB review is required
- eCFR: 21 CFR 312.62, investigator recordkeeping and record retention
- eCFR: 21 CFR 312.57, sponsor recordkeeping and record retention
- eCFR: 21 CFR 812.140, records (device investigations)
- eCFR: 21 CFR 11.1, scope of Part 11
- eCFR: 42 CFR 11.22, which applicable clinical trials must be registered
- eCFR: 42 CFR 11.48, clinical trial results information
- eCFR: 42 CFR 11.10, definitions (applicable clinical trial)
- eCFR: 45 CFR 46.104, exempt research
- eCFR: 45 CFR 46.116, informed consent and broad consent
- eCFR: 28 CFR part 202, access to US sensitive personal data by countries of concern
- EUR-Lex: Regulation (EU) 2016/679 (GDPR)
- EUR-Lex: Regulation (EU) No 536/2014 on clinical trials
- EDPB: Opinion 3/2019 on the interplay between the Clinical Trials Regulation and the GDPR (PDF)
- Vivli: Data Use Agreement v1.4 (PDF)
- The YODA Project: Data Use Agreement
- Reginfo.gov: HHS Unified Agenda entry RIN 0937-AA16 (Common Rule revisions)
