Opinion for the Court filed by Circuit Judge LEVENTHAL.
Concurring opinion filed by Circuit Judge MacKINNON.
Dissenting opinion filed by Circuit Judge BAZELON.
LEVENTHAL, Circuit Judge:
In its broad aspect this appeal presents the question whether and under what conditions data compiled by a private group that is receiving money under a federal grant-in-aid program are or become “agency records” by virtue of the fact that the agency has funded the program and has the authority to demand those records.
An action was brought by specialists in the treatment of diabetes, as individuals and a committee,1 to obtain raw research data of the University Group Diabetes Program (UGDP). The program is a privately conducted and federally funded long-term clinical study of the treatment of diabetes, that has reported certain harmful consequences attendant upon the long-term use of oral hypoglycemic agents. Plaintiffs question the validity of the study, and are concerned lest a useful therapeutic tool be unnecessarily removed from the market. They seek access to the raw data in order to implement their challenge to the study’s validity.
The action was brought under the Freedom of Information Act (FOIA), 5 U.S.C. § 552. That Act is addressed to each “agency” of the Federal Government as defined.2 Broadly speaking, and subject to [1131]*1131exceptions, it directs each agency to make available to the public certain information, and also “agency records.” It establishes the District Court’s “jurisdiction to enjoin the agency from withholding agency records and to order the production of any agency records improperly withheld from the complainant.” 5 U.S.C. § 552(a)(4)(B).
A. BACKGROUND
1. The UGDP Study and the Sponsoring Institute
The UGDP is a study funded by 13 federal grants administered by the National Institute of Arthritis, Metabolism and Digestive Diseases (hereafter sometimes Institute or NIAMDD). That institute is an “agency” under the Act, being part of the National Institutes of Health, which in turn is an organization within the Public Health Service, in the Department of Health, Education and Welfare. The grants were made under the statutory grant-in-aid authority of the Public Health Service Act, 42 U.S.C. § 241(c). The grants were made to each of 12 participating university medical centers on the basis of their applications, and a grant was made to the UGDP Coordinating Center at the University of Maryland.3
The pertinent background facts are presented in the affidavit of Dr. G. Donald Whedon, Director of NIAMDD:
4. The inspiration for the UGDP study came from private non-government physicians and scientists in mid-1959. Between 1959 and 1961, before the study actually began with the entry of the first patients, the design, methods, and objectives of the study were evaluated by persons associated with the UGDP and representatives of NIAMDD. The Food and Drug Administration was not involved in the planning, inception, or design of the UGDP study. The study was funded by NIAMDD as part of its responsibility to support research in the field of diabetes and not with any specific regulatory objective in mind.
sf: * * * * *
9. The UGDP raw data (e. g., patient charts and forms) are the property of the individual investigators and the Coordinating Center and are not owned by NIAMDD. Furthermore, it is not the normal practice of NIH or this Institute to require grantees to submit their raw data for review and, in fact, submission of raw data to the institute is extremely rare. Management of the day-to-day operations of grant-supported activities is the responsibility of the grantee. Supervision of the grantee’s funded activities by this Institute is generally limited to review of periodic reports submitted by the grantee. (45 CFR §§ 74.80, 74.82). Due to the large number of research grants outstanding — currently approximately 1800 — it would not be physically possible for the Institute to subject raw data, if submitted, to critical review, and' to require submission of the raw data of the UGDP study would have been an extraordinary requirement. It is the practice to evaluate applications for renewal grants on the basis of progress reports and final reports submitted to NIH. This practice was followed with respect to the UGDP grants. No specific [1132]*1132provisions of the UGDP grants required submission of raw data to the Department of Health, Education and Welfare. Pursuant to 45 CFR § 74.23, officers or employees of the Department could obtain access to the raw data for purposes of audit inspection and copying if access is deemed pertinent to the grant. The raw data which are the subject of this case have never been seen by, or been in the possession of, any officer or employee of the National Institutes of Health. * *
The particular documents sought by the plaintiffs in this case are observations on over 1000 diabetic patients, who were monitored from 5 to 8 years. It is estimated that there are some 55 million such documents.
In June, 1970, the UGDP investigators made a presentation of the methods and initial results of their study at the annual meeting of the American Diabetes Association. The results indicated that the administration of tolbutamide (an oral hypoglycemic drug) to mild adult-onset diabetics led to a death rate from cardiovascular disease higher than that of groups treated with diet alone, with a fixed dosage of insulin, or with a variable dosage of insulin. The findings were published in the December 1970 Journal of the American Diabetes Association. During 1970 and 1971, over a dozen articles were published in medical journals concerning the study, some supportive and some critical.4
The NIAMDD contracted in 1972 with the Biometric Society, a private international professional society of biostatisticians, for an in-depth assessment of the quality of the UGDP study. The Society made a report to the Institute in 1974 that apparently found some merit on both sides of the controversy. It concluded that while some of the criticisms of the UGDP study were valid most were unpersuasive, and the evidence of harmfulness adduced in the UGDP study was “moderately strong.” This was made public in the American Medical Association Journal for February 1975.5
2. Food and Drug Administration
The Food and Drug Administration of HEW, on being apprised of the UGDP results, issued in its October, 1970, Bulletin to the medical community a recommendation that tolbutamide should be used only in cases of adult-onset, stable diabetes that could not be controlled by diet and could not be treated with insulin. A June, 1971, FDA bulletin proposed changes in labeling of oral hypoglycemic drugs to warn of cardiovascular hazards. Plaintiff committee sued to enjoin the proposed labeling on ground of deficiencies in the UGDP study, and the First Circuit remanded to the FDA for exhaustion of administrative remedies.6
The FDA deferred further action on the labeling pending the review of the UGDP study by the Biometric Society. As already noted, the 1974 report of the Biometric Society was mixed, but overall found “moderately strong” evidence of harmfulness in the UGDP study. Its contract with NIAMDD did not require the Society to seek access to the UGDP raw data, but it apparently did examine some of the raw data.7 The contract did not require the [1133]*1133Society to submit any raw data to the Institute, and none was submitted.
3. FOIA requests and District Court proceedings
Stressing that the raw data had been made available to the Biometric Society, plaintiffs’ committee began a series of FOIA requests in 1974 and 1975 for access to the raw data and a copy of the draft report of the Biometric Society. Plaintiffs were given preliminary galley proofs of the report later published in the AMAJ. HEW notified plaintiffs on August 7, 1975, that the raw data were the property of those engaged in the UGDP study and had not been reviewed or even seen by either the UGDP sponsor (NIAMDD) or FDA.
This FOIA action was begun on September 30, 1975. The complaint sought the production of the raw data, defined as consisting of the forms transmitted to the Coordinating Center and the computer tapes and/or programs on the basis of which the data were analyzed. The complaint also sought a draft report of the Biometric Society.8
On Feb. 5,1976, the district court granted the motion of the HEW officials to dismiss the complaint, on the ground that no official or employee of HEW is now or has ever been in possession of the raw data relating to UGDP, that these raw data are the property of the individual investigators and UGDP study coordinating center, and in the Center’s possession, custody and control; that neither the investigators nor the Coordinating Center is an “agency” within 5 U.S.C. § 552, and that the raw data are not “agency records” subject to the disclosure provisions of FOIA.9
4. Developments pending appeal
On July 25, 1977, while the appeal to this court was pending, Secretary of HEW Cali-fano issued an imminent hazard order suspending new drug applications for phenfor-min (another oral hypoglycemic drug), and there ensued administrative withdrawal hearings. This court requested supplemental memoranda of the parties on the question of whether data that would become available to plaintiffs as a result of these administrative proceedings would moot the present controversy. The federal appellees put it that there is neither certainty nor likelihood that plaintiffs will obtain access to all the data they seek as a result of the phenformin proceeding. They note, for one thing, that phenformin was only one of the oral hypoglycemic drugs subject to the warning of the UGDP study, the principal one being tolbutamide.
However, it appears that the FDA did examine certain of the underlying raw data (a small portion, quantitatively) in the course of a recent limited audit of the UGDP, and that this portion of the underlying information (except patient-identifying information) has been made available to plaintiff-appellants, and to other interested persons, participating in the phenformin proceeding. The federal appellees’ memorandum states: “The FDA has no present intention of obtaining the remaining portions of the UGDP raw data through the auditing rights of the Secretary.”
B. ANALYSIS
We rule that the public at large does not have a right under the Freedom of Information Act to the underlying raw data in the hands of the investigators and university groups who conducted the UGDP study program of diabetes under grants from the federal government.
1. The plaintiffs are a respected group of medical specialists asserting that their access to the data would inure to the [1134]*1134public interest, by virtue of their concern that the use of drugs they deem valuable may be inhibited. We begin our analysis by observing that in this proceeding under the Freedom of Information Act, the court cannot give any weight to such a consideration.
The only claim ascertainable in this FOIA action is the right of any member of the public, motivated by whatever reasons. The Freedom of Information Act does not depend on a showing of need or interest by the particular applicant for the records. Any showing of need or interest is irrelevant.10 Such considerations as need, interest, or public interest may bear on the agency’s determination of the order of processing applications, but they have no bearing on the substantive right under FOIA to access to the document.11
2. To avoid any possible misunderstanding, we articulate that our ruling embodies no implication as to whether plaintiff physicians will have a right of access to the data underlying the UGDP study in connection with any existing or future actions of the Food and Drug Administration. That issue is distinctly different from what is before us now, and would have to be decided in the light of the record before the FDA.12
The FDA and NIAMDD are both in HEW, but that department is a conglomerate that embraces many and distinctly different activities. Insofar as it is engaged, through FDA, in a regulatory program, it may be subject to requirements of revelation that go beyond the FOIA’s rules that govern all agencies. The FDA’s regulatory actions are not before us in this FOIA lawsuit, which focuses on whether data become HEW records by virtue of study and granting activities (of NIAMDD).
[1135]*11353. This action requires that the persons invoking the FOIA show that they seek “agency records.” The NIAMDD is a government agency, of course. But the persons or institutions who receive study grants from that Institute, or indeed from any branch of the federal government, do not on that account become government agencies.
To some extent, our path is lighted by United States v. Orleans, 425 U.S. 807, 96 S.Ct. 1971, 48 L.Ed.2d 390 (1976). The case involved the Warren-Trumble council, a community action agency operating as a non-profit corporation under Ohio law, that was funded entirely by a federal agency, the Office of Economic Opportunity. Under the Economic Opportunity Act of 1964, the OEO furnished financial assistance to a community action agency, in turn defined as one designated by the state to plan and administer a community action program of “services and activities having a measurable and potentially major impact on causes of poverty in the community.” The issue was whether or not one of the activities of the Ohio community action agency, the sponsoring of recreational outings for children, if conducted negligently, could give rise to an action under the Federal Tort Claims Act. The Supreme Court held that it could not, since the council was not a federal agency or instrumentality, and its employees not federal employees. The Court found that a critical element in distinguishing a federal “agency” from either a contractor with the federal government or a grantee of the federal government, was the federal government’s “control [of] the detailed physical performance.”13
Our decision today is congruent with our decision in Washington Research Project v. HEW, 164 U.S.App.D.C. 169, 504 F.2d 238 (1974), which reversed a district court order granting disclosure of certain reports made to the National Institute of Mental Health, a unit of the Public Health Service of HEW. The case involved reports made, on applications for research support, by peer review groups (“initial review groups” or IRG). The IRG peer review system was established by the government to assure competent evaluation of proposals through the use of “expertise of nongovernmental consultants functioning in panels organized around particular specialized disciplines within the broader field of biomedicine.”14
The reports sought included a “site visit” to observe the pertinent experimental technique, and a “summary statement” of the observations and deliberations of the group, prepared by a NIMH staff member assigned to assist the group. The legal issue focused on whether the initial review group was itself a government “agency,” in which case its own reports would be “final opinions” required to be disclosed under FOIA, and not intra-agency memoranda excluded under exemption 5. Acknowledging the “myriad organizational arrangements for getting the business of the government done,”15 the court concluded that “the IRG’s are advisory committees, performing staff functions through the medium of outside consultancy, and are not agencies.”16 It observed, significantly, “Employing consultants to improve the quality of the work that is done cannot elevate the consultants to the status of the agency for which they work unless they become the functional [1136]*1136equivalent of the agency, making its decisions for it.” 17
4. Plaintiffs seek to avoid a head-on contention that federal grantees be assimilated as federal agencies. Instead they emphasize a congeries of considerations that they think cumulate to a right of public access to documents in the hands of the grantees.
In addition to the responsibility of plaintiffs and a claim of public interest in their access, which we have already shown to be irrelevant, plaintiffs stress the following; This was a multi-million dollar study, entirely funded by the federal government, of such a scale as to be non-replicable by private efforts and a unique public resource. By contract and regulation, the raw data underlying the study are available for government review, copying and storage. The government’s exercise of its rights of audit demonstrates its “complete dominion and control” over the data through the audit process.
The Institute’s grant documents establish its right of access to “any books, documents, papers, and records of the grantees” for certain purposes. To the extent that the language of the grant is material, it indicates that these are not agency records prior to the exercise of that right.
Plaintiffs’ claim is in effect an assertion that the federal government should be required — formally or constructively — to exercise its contract-grant right of access in order to provide general public access. We cannot accept this proposition. The Freedom of Information Act only gives a right of access to agency records in existence. It does not confer a right to have the government generate agency records, either by creation, subpoena or contract demand. That conclusion is implicit in NLRB v. Sears, Roebuck & Co., 421 U.S. 132, 95 S.Ct. 1504, 44 L.Ed.2d 29 (1975). The Court there granted the public a right to the production of the agency’s appeal memorandum, pursuant to its understanding that the Act “represents a strong congressional aversion to ‘secret [agency] law.’ ” (421 U.S. at 153, 95 S.Ct. at 1518). However the Court held that the public had no right to a judicial requirement that the agency produce or create explanatory material in the case of an appeals memorandum that referred only conclusorily to the “circumstances of the case.” See 421 U.S. at 161, 95 S.Ct. at 1521:
The Act does not compel agencies to write opinions in cases in which they would not otherwise be required to do so. It only requires disclosure of certain documents which the law requires the agency to prepare or which the agency has decided for its own reasons to create.
The governing principle is that only if a federal agency has created or obtained a record (or has a duty to obtain the record)18 in the course of doing its work, is there an agency record that can be demanded under FOIA.19
[1137]*11375. Overarching policy considerations are stressed by physician applicants. There is a plea for liberal reading of reform legislation. We agree that this reform legislation should not be niggardly construed in contravention of legislative objective. The “basic thrust” of the Act embraces “a general philosophy of full agency disclosure” subject to specific exemptions and the objective “to pierce the veil of administrative secrecy and to open agency action to the light of public scrutiny.”20
However, the general policy of avoiding agency secrecy does not give a charter for extending the law beyond the domain of “agency” and “agency records” that is the keystone of the Act. To stretch for data in the possession of federal grantees, cannot be justified as within the fair contemplation of Congress either at the time the law was passed or amended, or even today under a doctrine of trying to reconstruct specific legislative intent in the light of the broad purposes disclosed by Congress.21
It is tautology to say that requiring disclosure of grantee records will promote the disclosure policies of FOIA. But disclosure is not required by the statute unless those records are agency records. Congress struck a balance in fashioning the FOIA, which precludes the boundless pursuit of one policy goal, even a dominant policy, to the exclusion of all countervailing considerations.
If the statute is to be given the kind of interpretation sought by plaintiff physicians, the impact would be far-reaching. The number of documents in any one study would be stupendous — reaching millions in the single ease before us. The number of federal grants and contracts is not a matter of record, but as was noted in Orleans, they account annually for disbursements in the billions. The awesome implications of plaintiffs’ contention cannot be shrugged off because modern technology permits access to documents on tape through computer printouts, without need for physical production.
Scientists engaged in research on federal grants must accept the fact that any documents filed with the federal government, whether on the scientists’ own initiative or an audit or other lawful demand, are subject to FOIA. Even in scientific terms, any such audit provides a surrogate for the kind of reliability usually accorded to scientific studies by replication of experiments when feasible. However, an undertaking to be audited by responsible personnel is not the same as an agreement to accept rummaging by the world at large.
The court will not trim FOIA by speculation as to adverse motivation or reaction of the scientists.22 Similarly, the court cannot supply the extension of the reach of the Act sought by plaintiffs by building on a policy speculation that such an extension would not throttle scientific cooperation and research. This involves matters beyond our proper sphere of judicial notice.
[1138]*1138What is requested in this action, in our view, is an extension of the statute on claim of public interest that must be appraised by the legislature which can give the subject extended study, elicit opinions from all interested sources, and consider the pro’s and con’s.
6. It is fitting to close by referring to the need, in any pondering of such extension of the FOIA, for considering the impact on the philosophy and purpose of Federal grant programs.
Grant programs represent a means for the governance of our society which is rooted in a pluralistic conception of the value of drawing on both private and governmental sources. A leading student of Federal grant law puts it23
The grant is assistance to an autonomous grantee. The grantee is not an arm, agent or instrumentality of the grantor. The employees of the grantee are not federal employees. The torts of the grantee are not federal torts. The property of the grantee is not federal property.
The reference to “an autonomous grantee” is a core concept, not an incidental observation. In a grant program the federal government gets the advantage of services rendered by someone who is doing his own thing, his own autonomous thing. It is not the same as a government operation in disguise.
Through its grants to university groups, the government obtains the efforts of creative persons who flourish in an academic atmosphere. Such arrangements provide a measure of detachment and independence from the mission of the government agency. The researchers may feel the tug of government purse strings, but they also feel answerable to the standards of their academic colleagues.
Plaintiffs cite the multi-million dollar nature of the study as a reason for access. There is at least a question whether the federal government could have conducted directly, through its own employees and resources, a study program so long in time, so broad in space, and covering so many patients and controls. Even in a case where the grant is to conduct a study that might conceivably be conducted by federal employees, there is an advantage in terms of effective government and advancement of the public interest if the study is done by various institutions. The government goes beyond the capabilities of its own employees, adding the spirit and insights of the scientists and students who have selected a different life style, at a center of learning.
As earlier noted, we are not concerned here with the kind of case where the federal government exercises detailed control over operations. Such a condition presents different considerations, as noted in Orleans. Nor do we have a suggestion of subterfuge, with a federal agency seeking to conduct research outside the scrutiny of government laws, by using facilities that are independent only nominally. The case before us concerns a UGDP study conceived in 1959 by private, non-government physicians and scientists. They developed their own methodology; it was not dictated by the federal government.
Of course, in any program funded by the federal government there is an opportunity for the government to assess the results of the performance and of any studies. There may also be directions by the federal government in certain matters of public policy that are essentially peripheral to the core of the work done. There may, for example, be a requirement of avoidance of discrimination on grounds of race, religion, creed or sex. There may be achievement of other government objectives which apply across the board to all activities financed by the federal government.
The central question is whether the government is really involved in the core of the program. At least in a case such as the one before us, where there was no claim of [1139]*1139significant government control of day-by-day operation, or detailed involvement in the planning or execution of the program, the overall concept of autonomy of grantees persists, even though there are federal objectives, right of federal audit and perhaps some overarching federal requirements.
At least a fleeting reference should be made to acknowledge that some of the federal grantees are institutions of the state governments.24 There are thus considerations of federalism involved. These are not necessarily of constitutional dimension. However, they are not without relevance in appraising the extent to which such grantees are automatically governed by rules provided by Congress for the federal agencies, such as govern access to records and meetings, or personnel management,25 or any other rules.
The foregoing matters indicate that a balance must be struck, one that considers the advantages of grantees that are autonomous and have value because they are not governmental, and the possibly conflicting policy that cherishes full and free public access to government agencies and shuns secrecy as invidious. Such a balancing is a task for the legislature. The extension of access sought by plaintiffs on the ground of public interest is not properly addressed to the courts.
Affirmed.