GENERAL ELECTRIC COMPANY, Plaintiff-Appellant, v. UNITED STATES NUCLEAR REGULATORY COMMISSION, et al., Defendants-Appellees, 750 F.2d 1394

POSNER, Circuit Judge. The Nuclear Regulatory Commission decided to make public a set of internal documents of the General Electric Company known as the “Reed Report,” and General Electric brought this suit to enjoin the Commission from doing so. The district judge upheld the Commission’s decision and General Electric has appealed, raising a number of questions, substantive and procedural, under a variety of statutes and regulations. The Reed Report is a compilation, made in 1975, of several hundred pages of studies of General Electric’s “boiling water” nuclear reactor, which is used to generate electrical power.

The report contains a number of criticisms of the design of the reactor, including criticisms of its safety. The existence of the report became known when employees of General Electric mentioned it in congressional hearings in 1976 and 1977 on the safety of nuclear reactors. Although the report or parts of it were shown to certain government officials around the same time, General Electric did not make the report public.

Then in 1978, in a proceeding before the Commission’s Atomic Safety and Licensing Board on whether to license a nuclear electrical generating plant called “Black Fox” which was to be powered by General Electric boiling-water reactors, an intervenor asked to be allowed to use the Reed Report to cross-examine some witnesses from General Electric. The Board issued a subpoena to General Electric to produce the report.

General Electric moved the Board to quash the subpoena; but before the motion was acted on, the controversy over the subpoena was settled by the company’s agreeing to excerpt parts of the report for the lawyers (and their technical consultants) in the licensing proceeding, subject to a protective order, and to make the whole report available to the Board in confidence. The Board used the report to frame questions that were asked at in camera hearings in the licensing proceeding.

While that proceeding was moving along in the usual leisurely fashion, the Commission was being asked, mainly by citizens’ groups hostile to nuclear power, to release the Reed Report under the Freedom of Information Act, 5 U.S.C. § 552. General Electric not only opposed these requests but asked the Commission to withhold the report under a regulation (“Rule 790”) which provides that if the Commission determines that a document submitted to it “contains trade secrets or privileged or confidential commercial or financial information,” it can withhold the document from the public. 10 C.F.R. § 2.790(b)(5).

The Commission asked a staff economist, Messier, to review the Reed Report to see whether it contained material that should be withheld. Messier concluded in a brief memo that it did not. General Electric had earlier submitted contrary affidavits. The Commission handed down its decision on October 9,1980. In a brief opinion, the Commission stated that the Reed Report was an “agency record,” and therefore within the scope of the Freedom of Information Act. See 5 U.S.C. § 552(a)(3), (4)(B).

On whether the report contained any confidential information, the Commission said that “the staff [had] advised the Commission that it did not have an adequate basis to conclude that release of the Report would cause substantial harm to GE’s competitive position,” and that the Commission — which at the time had only four members — was evenly split on whether disclosure of the report would make it harder for the Commission to obtain similar information in the future. Nevertheless the decision was to release the report. The next day, General Electric moved the Commission under 10 C.F.R. § 2.790(

c) to return the Reed Report to the company. This

section of the regulation provides (with an immaterial exception) that the Commission will give anyone whose request to withhold a document under

section 2.790(

b) is denied 30 days to withdraw the document before it is made public. Nevertheless the Commission summarily denied General Electric’s request; the Commission considers subsection (

c) inoperative if a Freedom of Information Act request is pending when the request to withdraw is made. The company then brought this suit in federal district court. In 1983, while the suit was pending, the application to license “Black Fox” was withdrawn (see Note, Public Utilities: The Black Fox Nuclear Project Cancellation Dilemma: Of Judicial Review and Reform of Oklahoma’s Administrative Process, 36 Okla.L.Rev. 190, 192-94 (1983)), ending the licensing proceeding without a decision on the merits.

The Commission’s terse orders — the first denying in a brief opinion General Electric’s request to withhold the Reed Report from public inspection and granting the Freedom of Information Act requests, the second denying without opinion General Electric’s request to withdraw the report— do not explicate the Commission’s grounds. The key to understanding them (so far as it is possible to do so) lies in the interplay between the Freedom of Information Act and Rule 790. The Act contemplates that each agency subject to it will promulgate regulations implementing the Act. See 5 U.S.C. § 552(a)(3).

The Commission has done this in 10 C.F.R. §§ 9.3-9.16 (see id. §§ 9.1-9.1a), but these regulations are not involved in this case. Rule 790 is an earlier regulation dealing with public access to Commission records. Its history is recounted in Westinghouse Electric Corp. v. NRC, 555 F.2d 82, 85-93 (3d Cir.1977). It long predates the Freedom of Information Act but was overhauled some time after the Act was passed, in order to conform to the Act, see 41 Fed.Reg. 11808, 11809 (March 22, 1976), and in its present form it tracks the Act closely.

The Commission seems, indeed, to consider the rule and the Act substantively identical. See (besides the decision under review, where that view is implicit) 37 Fed.Reg. 15127 (July 28, 1972); Wisconsin Electric Power Co., 15 N.R.C. 281, 287 (Atomic Safety & Licensing Bd. 1982); Pacific Gas & Elec. Co., 5 N.R.C. 1398, 1402 n. 10 (Atomic Safety & Licensing Bd. 1977); Kansas Gas & Electric Co., 3 N.R.C. 408, 415 (Atomic Safety & Licensing Bd. 1976). The rule could not grant the public less access to the Commission’s records than the Act does, because the Commission admits to being fully subject to the Act.

But the rule could be broader. The Atomic Energy Act — the Commission’s organic statute — could authorize or even require the Commission to disclose more information than the Freedom of Information Act requires it to disclose. Hence the need perceived by the Commission (see 39 Fed.Reg. 40960, 40961 (Nov. 22, 1974)) for a regulation such as Rule 790 that establishes a policy on disclosure, as distinct from merely implementing the Freedom of Information Act — a statute that requires some, but as we shall see does not forbid any, disclosure.

Only agency records must be disclosed either under the Freedom of Information Act or under Rule 790(a), which differs in this respect only in substituting, apparently without intending a substantive change, see Kansas Gas & Elec. Co., supra, 3 N.R.C. at 415, the words “final NRC records and documents” for the statutory term “agency ... records," and which, as we have said, could not be narrower than the Act. So when faced with requests under the Act for the Reed Report, the Commission had first to decide whether the report was indeed an agency record.

Having decided it was, the Commission had next to consider whether the report might be exempt from disclosure.

The only exemption the Commission discussed was the one for “trade secrets and commercial or financial information obtained from a person and privileged or confidential,” which appears in the Act at 5 U.S.C. § 552(b)(4), and, in the identical words, in the regulation at 10 C.F.R. § 2.790(a)(4), being known in both places as “exemption 4.” Read literally, exemption 4 would shield virtually every document that a company chose not to make public; but the cases interpreting the provision have narrowed it considerably by holding that information that is not a traditional type of trade secret (of the secret-formula variety) is within exemption 4 only if disclosure would either inflict substantial competitive harm on the owner of the information or make it difficult for the agency to induce people to submit similar information to it in the future.

See, e.g., National Parks & Conservation Ass’n v. Morton, 498 F.2d 765, 770 (D.C.Cir.1974). These were the two criteria that the Commission addressed in its opinion of October 9, 1980. The first appears in Rule 790 itself as whether public disclosure “is likely to cause substantial harm to the competitive position of the owner of the information.” 10 C.F.R. § 2.790(b)(4)(v). The second (effect on information flow to agency) does not appear in the regulation but the Commission uses it anyway — illustrating our earlier point about the convergence of Rule 790 with the Freedom of Information Act.

Although the Commission split evenly on whether the Reed Report met the second criterion, the consequence was to order disclosure. Why the tie should go to the requester and not to the owner of the information requested is not apparent to us, but as General Electric does not question the Commission’s practice, neither shall we. If the Commission had decided that the Reed Report was within exemption 4 and so had refused to release it, the groups that had requested its release under the Freedom of Information Act could have sued the Commission under the Act.

But since the Act is merely a disclosure statute and creates no legally protected right of secrecy, Chrysler Corp. v. Brown, 441 U.S. 281, 290-94, 99 S.Ct. 1705, 1711-14, 60 L.Ed.2d 208 (1979), General Electric cannot base this suit to prevent the release of the Reed Report on the Act. It can, however, complain that releasing the report would violate some other law, including a Commission regulation (Rule 790), or simply that the Commission’s decision was arbitrary or capricious. Either way, there would be a violation of the Administrative Procedure Act redressable in this injunctive action.

See 5 U.S.C. §§ 702-704, 706(2)(A), (C); 28 U.S.C. § 1331. General Electric has made both types of argument here (violation of law; arbitrary and capricious action). For the first type, it relies on Rule 790. Subsection (b)(5) requires the Commission, if it finds that a document contains confidential information, to determine “whether the information should be withheld from public disclosure pursuant to this paragraph.” 10 C.F.R. § 2.790(b)(5)(ii). But no criteria for that determination are set forth. True, (b)(5)(

i) says that if the document contains confidential information, the Commission will determine “whether the right of the public to be fully apprised as to the bases for and effects of the proposed action outweighs the demonstrated concern for protection of a competitive position.” But there is no commitment to withhold the document if the balance inclines in favor of protection, though maybe such a commitment is implicit.

In any event, since Rule 790 cannot be narrower than the Freedom of Information Act, information that is not confidential within the meaning of exemption 4 cannot be confidential within the meaning of Rule 790(b)(5). But the Freedom of Information Act and Rule 790 do not coincide in all respects. The last sentence of subsection (b)(5) requires the Commission to return a document to whoever submitted it if the document “is deemed by the Commission to be irrelevant or unnecessary to the performance of its functions.” This provision, however, is doubly inapplicable to this case.

Although the Black Fox proceeding is now over with, it was still in progress when General Electric made and the Commission denied its request to withdraw the Reed Report. And the provision seems directed to cases where the Commission decides not to use a document — not to cases where the Commission has used it and no longer needs it. It cannot be that the Commission must immediately divest itself of the record of a proceeding when the proceeding ends — let alone before. Subsection (

c) of Rule 790, also a provision without counterpart in the Freedom of Information Act, provides (so far as is pertinent here) that if the Commission denies an application to withhold information under subsection (b), the applicant may withdraw the document before it is released to the public, provided he makes a timely request for withdrawal, as was done here. Read literally, this would utterly defeat the application of the Freedom of Information Act to any document submitted by a private party. Whenever someone requested such a document under the Act, the owner would counter with a motion under 790(

b) to withhold the document from the public, and if that motion was denied he would then withdraw it from the agency under (c). So if the Black Fox reactor project had been licensed, at the end of the licensing proceeding everyone who had submitted documents in connection with it could have withdrawn the documents from the Commission, with the result that there might have been no public record underlying an important Commission action. To avoid such absurdities, which would certainly violate the spirit, and maybe even the letter, of the Freedom of Information Act, see Kissinger v.

Reporters Committee for Freedom of the Press, 445 U.S. 136, 155 n. 9, 100 S.Ct. 960, 971 n. 9, 63 L.Ed.2d 267 (1980); Weisberg v. U.S. Dept. of Justice, 705 F.2d 1344, 1363 (D.C. Cir.1983), the Commission reads Rule 790(

c) as not applicable once a Freedom of Information Act request is filed. This is a reasonable

interpretation of its own regulation. Apart from its complaints about alleged violations of the regulation, General Electric argues that the Commission acted unreasonably (the practical meaning of arbitrarily or capriciously) in deciding to release the Reed Report. And it is this argument that brings the Freedom of Information Act most directly into play in this case. Although as we have said General Electric’s suit could not be based on the Act, the only reason the Commission gave for releasing the Reed Report is that the Freedom of Information Act required it to do so.

The Commission never exercised an independent discretion to decide whether, if not required to release the report, it should do so anyway. Maybe this is because it interprets Rule 790(b)(5)(

i) as (implicitly) requiring it to withhold any information that is confidential within the meaning of the rule, which may be the same meaning as that of exemption 4 in the Freedom of Information Act. In any event, if the Commission was wrong in its

interpretation of the Act, the case would have to be remanded to it for an exercise of whatever discretion it may have under its rules to release documents whose release is not compelled by the Act. So we must decide whether the Commission interpreted and applied the Freedom of Information Act correctly; and this depends first of all on whether the Reed Report is an agency record, for if it was not, then the Commission was wrong to think the Freedom of Information Act required that it release the report.

Although the Act does not define the term “agency record,” there is no doubt that it covers not only documents created by the agency but also documents submitted to the agency for use in carrying out its duties. See, e.g., Weisberg v. U.S. Dept. of Justice, 631 F.2d 824, 827-28 (D.C.Cir. 1980). The purpose of the Act — to give the public access to information on which the government bases action, see, e.g., NLRB v. Robbins Tire & Rubber Co., 437 U.S. 214, 242, 98 S.Ct. 2311, 2326, 57 L.Ed.2d 159 (1978) — would be impeded if “agency record” excluded documents that had moved the government to act. See Weisberg v.

U.S. Dept. of Justice, supra, 631 F.2d at 828. So the record compiled in a nuclear licensing proceeding must be an “agency record,” and the Reed Report was a part of the record compiled in the Black Fox proceeding. It is irrelevant to this conclusion that the report was the subject of a protective order and was received in camera. See, e.g., Petkas v. Staats, 501 F.2d 887, 889-90 (D.C.Cir.1974).

The Freedom of Information Act would have little bite if government agencies could escape its provisions just by stamping documents “classified.” Since sophisticated corporations that deal extensively with the government know this, General Electric’s claim to have been misled into dropping its motion to quash the subpoena by a belief that the Commission had agreed not to disclose information received in confidence is suspect — especially given the Commission’s emphatic reservation of the right to disclose such information. See, e.g., 10 C.F.R. § 2.790(b)(2); Chem-Nuclear Systems, Inc., 9 N.R.C. 755, 758-59 (Exec.

Dir. for Operations 1979). It is also irrelevant that the licensing proceeding aborted after the Freedom of Information Act requests were made (several years after). Nothing in the Act suggests that a request for disclosure is unripe until there has been final governmental action. That would often be too late to do anyone any good. Moreover, the requests were not only made, but granted, before the proceeding aborted; and it cannot be right that a firm can prevent the disclosure of documents simply by spinning out its challenge to disclosure until the proceeding in which the agency used the documents has ended.

And it is irrelevant that the Commission made only limited use of the Reed Report in the licensing proceeding. For one thing, when the Freedom of Information Act requests were made, in the midst of the proceeding, it was unclear how much use the Commission would make of the report. For another, the public has a legitimate interest in verifying an agency’s claim that only part of a document was relevant to the agency proceeding and was used by the agency.

Although only excerpts from the Reed Report were placed in the formal record of the licensing proceeding, the Commission received the whole report in order to verify that the portions had been correctly excerpted; and the public has an interest in knowing whether the Commission (through its Atomic Safety and Licensing Board) did verify this — not necessarily an overriding interest, but interest enough to get over the statutory threshold.

A related point is that while the amount of the agency’s use of the document might be relevant to whether a document was within any of the Act’s exemptions, to make it a criterion of whether the document is an agency record would muddy what ought to be a clear threshold question. General Electric cites cases holding that neither an agency’s access to documents (that is, its unexercised power to obtain them) nor even the agency’s physical custody of documents (not created by the agency) is enough, in and of itself, to turn documents into agency records. See, e.g., Kissinger v.

Reporters Committee for Freedom of the Press, supra, 445 U.S. at 157, 100 S.Ct. at 972; Forsham v. Harris, 445 U.S. 169, 185-86, 100 S.Ct. 977, 986-87, 63 L.Ed.2d 293 (1980); Wolfe v. Department of Health & Human Services, 711 F.2d 1077, 1078-80 (D.C.Cir.1983). The good sense of these decisions is evident.

To bring under the Freedom of Information Act documents that an agency might have gotten hold of by subpoena or otherwise but didn’t, or that happened to be on the agency’s premises because an employee of the agency had carried it there — the morning newspaper, for example, or a personal diary — would make the Act a record-acquisition and record-retention act, which it is not. See, e.g., Kissinger v. Reporters Committee for Freedom of the Press, supra, 445 U.S. at 152, 100 S.Ct. at 969; Weisberg v. U.S. Dept. of Justice, supra, 705 F.2d at 1362; Yeager v.

Drug Enforcement Administration, 678 F.2d 315, 321 (D.C.Cir.1982). But the Reed Report was demanded and received by the agency for use in a formal proceeding. The next question is whether the Reed Report is within one of the Freedom of Information Act exemptions. The Commission discussed only exemption 4, but General Electric argues from 3 (“specifically exempted from disclosure by statute”) as well.

Though in part just a reprise of the company’s argument that the Commission violated a specific law (and was not merely arbitrary and capricious) in ordering the Reed Report made public, the argument makes two new points, which being pure issues of law we can decide without benefit of the Commission’s views expressed in a decision. The first is that the Atomic Energy Act “specifically exempted” the Reed Report from disclosure.

Section 103(b)(3) of that Act makes it a condition of obtaining a nuclear license that the applicant “agree to make available to the Commission such technical information and data concerning activities under such licenses as the Commission may determine necessary to promote the common defense and security and to protect the health and safety of the public. All such information may be used by the Commission only for the purposes of the common defense and security and to protect the health and safety of the public.” 42 U.S.C. § 2133(b)(3).

Although this language lacks specificity, a condition of invoking exemption 3, and does not even seem related to disclosure, it apparently was intended to express a “congressional policy which disfavors disclosure of proprietary information.” Westinghouse Elec. Corp. v. NRC, supra, 555 F.2d at 92; see also id. at 89-91.

Therefore it might require the Commission to withhold disclosure of an agency record, see id. at 88-93, quite apart from exemption 3, at least if the record was within exemption 4 (for then there could be no conflict between the Freedom of Information Act and the Atomic Energy Act if disclosure was withheld, and therefore no need to decide which statute should take precedence if they are in conflict). But the legislative history that General Electric cites makes clear that the purpose of

section 103(b)(3) was “to protect the property right, the commercial right, which a licensee as a developer of a new procedure, new idea, should properly have,” Hearings on S. 3323 and H.R. 8862 to Amend the Atomic Energy Act of 1946 Before the Jt.Comm. on Atomic Energy, 83d Cong., 2d Sess. 925 (1954) (remarks of Congressman Cole, the committee’s chairman) — not to prevent the disclosure of embarrassing, because self-critical, information. See also the other legislative-history materials cited in the Westinghouse case, supra, 555 F.2d at 90-91. In any event, we cannot find in

section 103(b)(3) a broader or more explicit directive to protect commercial secrecy than exemption 4 itself contains. General Electric also relies on the Trade Secrets Act, 18 U.S.C. § 1905. There is a difficult problem, well discussed in 1 Davis, Administrative Law Treatise § 5:31, at pp. 396-98 (1978), in integrating this statute with exemption 3: the Trade Secrets Act forbids disclosure of trade secrets “to any extent not authorized by law,” thus inviting reference to the Freedom of Information Act, a statute authorizing disclosure, but exemption 3 invites reference back to the Trade Secrets Act.

One can break the circle, however, by bringing in exemption 4. If a supposed trade secret is not protected by that exemption, the Freedom of Information Act requires its disclosure, and then the Trade Secrets Act by its own terms does not forbid it and exemption 3 does not come into play. If this is the right approach, the Trade Secrets Act has no independent force in cases where the Freedom of Information Act is involved, a conclusion strongly hinted at by the Supreme Court in a footnote in the Chrysler case. See 441 U.S. at 319 n. 49, 99 S.Ct. at 1726 n. 49.

Exemption 4 is broadly worded, and it is hard to believe that Congress wanted seekers after information to stub their toes on a rather obscure criminal statute almost certainly designed to protect that narrower category of trade secrets — secret formulas and the like — whose disclosure could be devastating to the owners and not just harmful. So if the Reed Report is not protected by exemption 4, even more clearly is it not protected by

section 1905 either. Turning to exemption 4, we understand General Electric to accept that the issue is whether releasing the Reed Report to the public would either make it hard for the Commission to obtain similar information in the future, or substantially harm General Electric’s competitive position; it quarrels only with the Commission’s negative answer. We have little difficulty with the Commission’s finding on the possible difficulty of getting like information submitted to it in the future, though this is the issue on which the Commission itself was split.

Recall that the Commission originally subpoenaed the report; though General Electric resisted the subpoena, the Commission may be confident of its ability to use its subpoena power to get all the information it needs to carry out its regulatory responsibilities, even though it must ask the Justice Department to enforce in court any subpoena not complied with voluntarily. See Washington Post Co. v. United States Dept. of Health & Human Services, 690 F.2d 252, 268 (D.C.Cir.1982).

The Commission may also believe that anyway it is in the driver’s seat in dealing with the manufacturers of nuclear reactors; that if they are not forthcoming with information about the safety problems of nuclear reactors their chances of getting licenses for nuclear reactors will be poor — and that they know this. And since the Commission is the body primarily hurt by shutting off the flow of information to it, we can assume that it lends a sympathetic ear to claims that disclosing a document will make it harder to get information in the future.

Thus, when it rejects such a claim, as in this case, there is no presumption that it is doing so out of an unreasoning hostility- Unfortunately, the Commission did not explain the basis for its action, leaving us to speculate along the foregoing lines; and in a different context that could be a very serious problem, as we are about to see.

But the main reason for a reviewing court’s forcing an agency to explain the basis of its decisions is to facilitate judicial review, and we do not think there is much room for judicial review of the quintessentially managerial judgment that disclosing information submitted to an agency in confidence will not prevent the agency from obtaining such information in the future. Especially where the Commission resolves doubts in favor of disclosure (unlike such cases as Washington Post Co. v.

United States Dept. of Health & Human Services, supra, 690 F.2d at 268-69, where the agency in question had ordered the information withheld from the public) — contrary to its own self-interest in promoting the flow of information to it — the chance of judicial reversal is so slight that it would be an idle gesture to require the Commission to devote resources to writing out a rationalization of its action. It is possible to overjudicialize the administrative process. The harder question is whether the Commission acted properly in turning down General Electric’s competitive-injury ground for invoking exemption 4.

Again our concerns are procedural rather than substantive. While General Electric’s competitors in the nuclear-reactor business would no doubt be delighted to be able to flag around to their customers a report in which General Electric criticizes its own reactor design, the competitive harm that attends any embarrassing disclosure is not the sort of thing that triggers exemption 4. There must be substantial competitive harm to the firm that owns the information sought to be made public, see, e.g., National Parks & Conservation Ass’n v.

Kleppe, 547 F.2d 673, 681-84 (D.C.Cir.1976), though of course the firm need not make its case of substantial competitive harm with anything like the rigor that would be demanded of a plaintiff in an antitrust suit, cf. Public Citizen Health Research Group v. FDA, 704 F.2d 1280, 1291 (D.C.Cir.1983). General Electric is the dominant producer of nuclear reactors; some of the criticisms contained in the Reed Report were revealed in the congressional hearings in which the existence of the report was disclosed; the report was already five years old when the Commission ordered it made public.

These things make it unlikely that the order will harm General Electric severely. Notwithstanding an oral request for the Reed Report by Westinghouse, which is General Electric’s principal competitor in the nuclear reactor business, the principal demanders of the report are opponents of nuclear power, who would like to embarrass the entire industry rather than shift business from General Electric to Westinghouse.

And though any disparagement of nuclear power could harm General Electric vis-a-vis producers of competing systems for generating electricity, the harm likely to be done by a single document in this respect is too speculative to bring exemption 4 into play. The problem with the Commission’s rejection of General Electric’s assertion of competitive injury is not that the rejection seems unreasonable but that the Commission did not write a reasoned decision.

It did not so much as acknowledge the existence of the affidavits that General Electric had submitted to show competitive injury; it did not mention letters in its files that suggested that the Reed Report was entitled to the protection of exemption 4; it did not even refer explicitly to the one-page memorandum from its staff economist, Messier, which was the only piece of evidence on the other side.

It merely stated a conclusion that General Electric had not shown substantial competitive injury — and stated this unclearly, saying only that “the staff [had] advised the Commission that it did not have an adequate basis to conclude that release of the Report would cause substantial harm to GE’s competitive position.” The Commission’s opinion on competitive injury, the whole of which we have just quoted, validates congressional criticisms of the excessive casualness displayed by some agencies in resolving disputes over the application of exemption 4. See H.R. Rep.

No. 1382, 95th Cong., 2d Sess. 2-3, 12, 25-26, 47-52, 61-62 (1978); S.Rep. No. 221, 98th Cong., 1st Sess. 14-19 (1983). True, a proceeding on a request for information is not required to be as elaborate as a licensing or other formal proceeding. And true, the principal affidavit that the company submitted was only two and a half pages long, and said only that the Reed Report would reveal to competitors General Electric’s “game plan” for improving its reactors and would enable the competitors to disparage its reactors to customers.

The competitors are not identified, and no attempt is made to assess the gravity of the competitive injury to General Electric. And true, Messier had made two valid points in his memo: that General Electric has very little competition in the nuclear-reactor business, and that if after five years the company still had not adopted the suggestions in the Reed Report for improving its reactors (if it had, it could not be hurt — not much anyway — by the release of the report), this would just be further evidence of how weak competition is in this industry.

But we do not even know whether this memo is the staff advice to which the Commission’s opinion refers. We can give no weight to Messier’s separate affidavit that was submitted to the district court. If the purpose had been to amplify the basis of the Commission’s action, well and good; a court reviewing informal agency action may receive evidence showing the reasons for that action. See, e.g., Clement v. SEC, 674 F.2d 641, 646 (7th Cir.1982). But while giving reasons why exemption 4 should not be applied, the affidavit does not attribute those reasons to the Commission as of the date of its decision.

If we relied on it in reviewing that decision, we would be violating the principle of SEC v. Chenery Corp., 318 U.S. 80, 94, 63 S.Ct. 454, 462, 87 L.Ed. 626 (1943), that an agency cannot defend its actions on grounds different from those on which the actions were based. See City Federal Savings & Loan Ass’n v. Federal Home Loan Bank Bd., 600 F.2d 681, 688-90 (7th Cir.1979).

Casting out the affidavit, however, highlights the Commission’s failure to give any reasons at all for accepting the staff’s rejection of General Electric’s claim of competitive injury, or even to indicate on what staff advice the Commission was relying.

Although it is unlikely that forcing the Commission to write another opinion will lead to a different result, especially since by the time it gets around to doing so the Reed Report will not be five years old, but ten years old (which means that General Electric will have won — maybe has won— this case by sheer litigation delay; the report is rapidly becoming an historical curiosity whose release can no longer harm the company), we cannot review an agency’s determination without knowing what it is based on; and like the Third Circuit in Chrysler Corp. v.

Schlesinger, 565 F.2d 1172, 1192 (3d Cir.1977) (vacated on other grounds under the name of Chrysler Corp. v. Brown, supra), we are unable to discover “the basis for the agency determination that the FOIA exemptions are inapplicable.” See also Camp v. Pitts, 411 U.S. 138, 142-43, 93 S.Ct. 1241, 1244, 36 L.Ed.2d 106 (1973) (per curiam); City of West Chicago v. NRC, 701 F.2d 632, 648-49 (7th Cir.1983); General Dynamics Corp. v. Marshall, 572 F.2d 1211, 1218 (8th Cir. 1978), remanded without opinion, 441 U.S. 919, 99 S.Ct. 2024, 60 L.Ed.2d 392 (1979), for further consideration in light of Chrysler v. Brown.

If, as with the Commission’s finding that the flow of information to it would not be impeded by releasing the Reed Report, a finding that General Electric’s competitive position would not be harmed substantially by the release could not be reviewed judicially because it was quintessentially judgmental, managerial, or discretionary, there would be little point in forcing the Commission to spell out the basis of the finding. But a finding on competitive injury is not of this nature.

It is akin to findings made all the time in antitrust and unfair-competition cases, and is effectively reviewable by appellate courts — provided the agency explains the finding in enough detail (which need not be extensive) to enable the court to perform its reviewing role; and the agency did not do that here. We emphasize the narrow scope of this ruling, however. We do not suggest that the Commission has to find that the release of the Reed Report at this late date would do substantial competitive harm to General Electric.

The memo that Messier prepared for the Commission may, in combination with all else that was before the Commission (or new evidence that the Commission may wish to gather on remand), fully justify its decision. And maybe General Electric’s own affidavits were so thin that the Commission could have rejected the claim of substantial competitive injury without considering the staff economist’s views.

But as we cannot reconstruct the Commission’s thinking process from an opinion that devotes one short sentence to the issue of competitive injury, we are forced to reverse the district court’s judgment and direct that court to remand the matter to the Commission for further (and we hope expedited) consideration of the limited respect in which we are unable to sustain the Commission’s action in ordering the Reed Report made public. Affirmed in Part, Reversed in Part, and Remanded.