Lawrence E. Wood appeals the denial by the district court of his motion for a new trial or dismissal of the charges against him, alleging the existence of newly-discovered evidence favorable to him that the government had failed to disclose. Wood’s appeal raises the question of the responsibility of the Food and Drug Administration (the FDA) to disclose in a criminal trial the content of Inves-tigational New Drug applications (INDs) that bear on the safety of the drug a defendant is charged with unlawfully dispensing. Holding that the FDA failed to discharge its obligation of disclosure, we remand to the district court for that court to determine what effect the FDA’s failure had on the outcome of the trial.

PROCEEDINGS

On June 18, 1991, Wood and his co-defendant Daniel R. Duchaine were charged in Count I with conspiracy in violation of 18 U.S.C. § 371 to defraud the FDA by obstructing the FDA’s function of ensuring that prescription drugs are safe and effective and dispensed pursuant to a prescription from a practitioner licensed by law to administer such drugs. The drugs in question were gamma hydroxybutrate and gamma hydroxy-butyric acid sodium salt (collectively GHB) and Clenbuterol, each alleged to be prescription drugs within the meaning of 21 U.S.C. § 353 (b)(1)(B). Counts Two through Six charged Wood with aiding and abetting Dú-chame in violation of 18 U.S.C. § 2 in distributing GHB and Clenbuterol without labelling that stated the manufacturer, distributor, and quantity.

Wood moved for discovery under Fed. R.Crim.P. 16. That rule provides in pertinent part:

Id. 16 (a)(1)(D). On December 9, 1991, the district court ordered the government to turn over to the defendant “all information falling within Rule 16, including information in the custody of the FDA, forthwith.”

At a hearing on December 16, 1991, the prosecutor told the court that in determining to seek the prosecution the FDA had relied on reports from various Poison Control testing and that those reports had been furnished to the defense. He added that there was information he had not turned over to the defense, information “concerning specific research done by companies on this drug, and it is protected under the Freedom of Information Act and Trade Secret Laws and the FDA lawyers have told me it cannot be turned over; and if it is turned over, then there has to be a very, very strict protective order. It is in archives and isn’t available for at least six to eight weeks even on an expedited basis.” The prosecutor added that the FDA had not relied on this information in deciding that GHB was a prohibited drug. The court responded: “If that’s all they relied upon to make the determination, that’s fine. You were ordered to provide whatever it is they relied upon in determining that this was a prohibited drug.” The defense did not ask for more.

After a jury trial, both defendants were convicted. Wood was sentenced to three months imprisonment, three months in a community correction center, three years of supervised release, a $3,000 fine and a $300 special assessment. We affirmed in an unpublished memorandum disposition of October 21, 1993. Wood’s petition for rehearing was denied and his suggestion for rehearing en banc was rejected on March 4, 1994. On April 18,1994, Wood moved for a new trial or dismissal of the charges. Meanwhile he served the confinement to which he had been sentenced.

The basis of Wood’s motion was the disclosure by the FDA in an entirely independent criminal trial in Arizona, United States v. Mark Thierman, of the existence within the FDA of the contents of certain INDs relative to GHB. In December 1992, the district court in Arizona ordered some of the IND materials disclosed to the Thierman defense. The material became public during Thier-man’s trial (which resulted in his conviction) in November 1993, and Thierman drew the material to Wood’s attention.

After Wood had made his April 19, 1994, motion for a new trial, the district court in his case considered the INDs from the Arizona trial, together with certain testimony in that trial. On June 11, 1994, the district court denied Wood’s motion. The district court did not give reasons for its ruling.

Wood appeals, arguing that the government failed to disclose at trial what Brady v. Maryland, 373 U.S. 83, 87, 83 S.Ct. 1194, 1196 , 10 L.Ed.2d 215 (1963) requires to be disclosed and what Fed.R.Crim.P. 16(a)(1)(D) and the district court’s order required to be disclosed. The government denies that the INDs were material to Wood’s defense and denies that it disobeyed the order of the district court.

ANALYSIS

First. The defense waived its right to the INDs under the Federal Rules of Criminal Procedure when it did not challenge the district court’s ruling on December 16, 1991. That ruling effectively narrowed the court’s earlier ruling of December 9, 1991, requiring the delivery by the government of all Rule 16 material in the custody of the FDA. The prosecutor did not hide from the court or the defendant that INDs existed and were not being disclosed. The defense acquiesced in the court’s limitation of discovery to the material on which the FDA had relied in deciding to treat GHB as a prescription drug. The prosecutor did not mislead the court and indeed in the course of his argument indicated that the INDs could be made available under a protective order, well before the trial, which was scheduled for April 1992, took place. When the defense at such a hearing held to consider the government’s compliance with Rule 16 was informed of the existence of the INDs and did not ask for them, it waived the right to have them produced pursuant to the court’s order and Rule 16.

Second. The INDs were Brady material, which the government had a duty to disclose; failure to do so denied Wood due process of law. Brady v. Maryland, 373 U.S. 83 at 87 , 83 S.Ct. 1194 at 1196 , 10 L.Ed.2d 215 (1963). This constitutional duty, established by the Supreme Court’s interpretation of our Constitution, was affirmative, anterior to, and weightier than the district court’s order and Rule 16. The duty arose from the unfairness of the government prosecuting a person on a criminal charge while keeping as a secret to itself materials that might exonerate him. It is a duty that assumes that the government’s aim in a criminal trial is not victory but justice. It is a duty, converted into a constitutional obligation, that could not be waived by a defendant in ignorance of what he was waiving, Brewer v. Williams, 430 U.S. 387, 404 , 97 S.Ct. 1232, 1242 , 51 L.Ed.2d 424 (1976); and so it was not waived by Wood here when he acquiesced in the narrowing of the disclosure order but was uninformed of the information that the government was withholding.

On December 16, 1991, the prosecutor did not know more of the contents of the INDs than Wood. But the FDA did. For Brady purposes, the FDA and the prosecutor were one. We need not decide how far the unity of the government extends under the Brady rule. We hold only that under Brady the agency charged with administration of the statute, which has consulted with the prosecutor in the steps leading to prosecution, is to be considered as part of the prosecution in determining what information must be made available to the defendant charged with violation of the statute. The government cannot with its right hand say it has nothing while its left hand holds what is of value. United States v. Endicott, 869 F.2d 452, 455 (9th Cir.1989); cf. Santobello v. New York, 404 U.S. 257, 262 , 92 S.Ct. 495, 498 , 30 L.Ed.2d 427 (1971). The government in the form of the prosecutor cannot tell the court that there is nothing more to disclose while the agency interested in the prosecution holds in its files information favorable to the defendant.

That the FDA files contained such information became apparent when the contents of the INDs, not their mere existence, was disclosed to Wood. To appreciate the significance of what became available, some review of the testimony at trial is helpful.

In order to prove what Count I charged, that GHB was a prescription drug, the government presented as a principal witness Dr. Wallace Winters, regional medical officer of the FDA in San Francisco. Winters testified to his experiments with administering GHB, to cats. He compared the toxic effects of GHB on cats as shown in his experiments to the effects of pheneylidine (PCP) and of LSD, both notorious hallucinogenic drugs; he testified that from the behavior of the cats subjected to GHB it could be inferred that they were suffering hallucinations. Although he testified that he had not experimented on human beings, the clear and powerful thrust of his testimony was that GHB was a dangerous drug whose hallucinogenic properties could affect humans who ingested it.

To the contrary, a fair amount of the material in the INDs showed that GHB, if properly taken by humans, was not dangerous to them. For example, one report states:

These reports would have been useful in impeaching Winters’ testimony. “[Ijmpeaehment evidence, as well as exculpatory evidence, falls within the Brady rule.” United States v. Bagley, 473 U.S. 667, 676 , 105 S.Ct. 3375, 3380 , 87 L.Ed.2d 481 (1985).

The government argues that disclosure was forbidden by 18 U.S.C. § 1905 . This statute applies to all employees of the United States, forbidding them to disclose confidential information in their possession “unless authorized by law.” It is difficult to see what could be more explicit authorization by law than the order of the district court enjoining on the government its constitutional duty to make Brady material available to the defense.

Third. Disclosure of the INDs would have had some impact on the GHB aspects of Wood’s trial. The impact the INDs would have had is not easy to determine. The story they told had been presented to the jury by the defense’s expert, Andrew Thomas Weil. Dr. Weil is a physician whose expertise in research is in psychoactive drugs, addiction, plant medicine, natural medicine, and alternative medicare. For thirteen years he had been a research associate at Harvard University in the field of ethnophar-macology or the study of drugs in different cultures; and he is on the editorial boards of the Journal of Ethnopharmacology and the Journal of Psychoactive Drugs. He had read over seventy published articles on GHB, had interviewed thirty persons who had used GHB, and had directed about a dozen patients in the use of GHB. Weil testified that the consensus in the scientific literature on GHB was that GHB “is an outstandingly safe pharmacological agent.” Weil was asked, “Is it an hallucinogenic drug?” He replied, “Absolutely not.” In further testimony he stated that GHB “has no resemblance to an hallucinogenic drug_ It is by chemical definition not an hallucinogenic drug.” Specifically asked about the published research of Wallace Winters, Weil described him as “out of the main stream,” doing research on cats not humans, using “enormously toxic doses” in his experiments, and employing a hypothesis as to the states produced by drugs that was Winters’ “private theory” not in accord with that of other researchers. Winter’s comparison of GHB with POP and LSD Weil described as confusing and unfounded. Weil testified that hallucinations could not be produced in animals as Winters had claimed, because “hallucinations are a product of the cortex and you can’t find that in animals.” Weil’s testimony, in short, if believed by the jury, would have thoroughly discredited Winters and made the IND material superfluous.

It is difficult for an appellate court to assess the degree to which the INDs would have appeared as chiefly cumulative to Weil’s testimony and the degree to which they would have added important corroboration to his testimony by being sources not suspect of bias for the defendant. That assessment of the materiality of the INDs belongs in the first instance to the court that conducted the trial. Holding that the INDs were Brady material that should have been provided by the prosecutor, we remand to the district court to make this assessment.

Fourth. Wood was convicted of conspiracy not only in regard to GHB but in regard to Clenbuterol. It is possible that Wood’s conviction could be sustained on the basis of an agreement by him with Duchaine to frustrate the function of the FDA in regulating Clen-buterol; if so, the withholding of the INDs as to GHB would have had an immaterial effect on Wood’s conviction. It is possible, but at this distance from the trial it does not appear likely for two reasons: (1) There appears to be very little evidence showing an agreement between Wood and Duchaine as to this drug; (2) Dr. Jean L. Fourcroy, a medical officer of the FDA, who appeared as a government expert on the side effects of Clenbuterol testified to experiments with the drug on humans in this country for a purpose that she “was not at liberty to say.” Reporter’s Transcript 315. Her reference may be to INDs; if that is the case, the government may have a Brady problem as to them. Again, remand to the district court is the route by which the significance of the evidence connecting Wood to a Clenbuterol conspiracy may be weighed and any Brady question in regard to this issue may be explored.

Fifth. That Wood has been convicted, lost his direct appeal, and served his sentence of confinement does not affect his right to seek a new trial or even dismissal of the charges. Justice done late is better than justice not done at all. But we do not prejudge the outcome of the remand.

The case is REMANDED to the district court for it to make findings of fact and state conclusions of law as to the materiality of the government’s Brady violations.