This case requires this Circuit to decide for the first time whether the so-called "military contractor defense" is available to shield a private defense contractor from liability to a serviceman killed as a result of an alleged design defect in the contractor's product. We hold that this Circuit will recognize such a defense, but we frame that defense and the rationales behind it somewhat differently from other circuits that have considered this question. Our analysis of the instant case under the terms of this defense leads us to AFFIRM the decision of the court below.

On December 12, 1979, at about 4 a.m., Navy Lt. J.G. Gary S. Shaw was killed when the plane he piloted, a Grumman A-6 (in its "tanker" version, the KA-6D), crashed into the ocean 100 miles off the coast of San Diego just two or three seconds after it was launched from the aircraft carrier Constellation. Although they recovered neither the body nor the wreckage, Navy investigators concluded that the most likely cause of the crash was the loss or failure of a bolt in the "stabilizer actuation system" or "longitudinal flight control system"1 of the aircraft, which instantly sent the plane out of control on launch. The A-6 carries no device to warn of such a failure and no back-up system to take over if this sort of breakdown occurs. Investigators ruled out pilot error as a cause of the accident.

The A-6 is the mainstay of the carrier-based Naval air attack fleet. Since Grumman Aerospace began to manufacture this plane for the Navy in the 1960's, six other accidents involving longitudinal flight control failures have occurred. Grumman advised the Navy to address this problem by installing so-called "self-retaining bolts"2 in the stabilizer system. This the Navy did in 1978, after a delay of some years for "fiscal reasons." Lt. J.G. Shaw's plane apparently crashed with one of the new bolts in place.

Edwin Lees Shaw, father of Lt. J.G. Shaw and personal representative of his estate, filed this wrongful death action on October 16, 1981, under the Death on the High Seas Act, 46 U.S.C.A. Sec. 761, et seq., and federal admiralty law alleging strict liability, negligence and breach of warranty against Grumman Aerospace. Shaw claimed that Grumman's A-6 design was defective because it failed to include any warning or back-up systems in the foreseeable event of a stabilizer control failure. In answer, Grumman asserted, inter alia, its immunity from suit under the "government contractor defense."

After a non-jury trial, the United States District Court for the Southern District of Florida, Aronovitz, J., entered judgment for plaintiff Shaw and awarded damages of $840,556.75 against Grumman, 593 F. Supp. 1066 (1984). In reaching its decision, the lower court recognized the form of the government contractor defense set out by the Ninth Circuit in McKay v. Rockwell International Corp., 704 F.2d 444 (9th Cir.1983), cert. denied, 464 U.S. 1043, 104 S. Ct. 711, 79 L. Ed. 2d 175 (1984), but found that Grumman did not prove the elements of that defense.

The Supreme Court has not addressed the issue of immunity from liability for military contractors who design products for government use. Four circuits have recognized varying forms of a military contractor defense.3 See Bynum v. FMC Corp., 770 F.2d 556, 561-62 (5th Cir.1985); Tillett v. J.I. Case Co., 756 F.2d 591, 596-598 (7th Cir.1985); Koutsoubos v. Boeing Vertol, 755 F.2d 352, 354-355 (3d Cir.1985); McKay, 704 F.2d at 448-451. Two other related defenses, which we refer to here as the "contract specifications defense" and the "government agency defense," have appeared from time to time in this Circuit. Discussion of these theories and the relationship among them has been frequent in recent years, but not invariably helpful. See generally, Bynum; Johnston v. United States, 568 F. Supp. 351 (D.Kan.1983).

The three defenses above are based on different theories of liability and require different elements of proof.

The "contract specifications defense" is in fact not, as the Bynum court points out, strictly speaking a defense at all. Bynum, at 560. When a contractor manufactures products to the order of another party, be that party the government or a private entity, he or she is simply not held to the same standard of care as an actual designer. Thus, the contractor will not be liable for damages caused by the product's design unless the specifications provided on order were so clearly defective and dangerous that a reasonably prudent contractor "would realize that there was a grave chance that his product would be dangerously unsafe." Restatement (Second) of Torts Sec. 404 comment a. (1965). See also Bynum, at 560; Johnston, 568 F. Supp. at 354. This higher threshold of fault for a contractor is obviously available only where fault is at issue--that is, in a suit brought under negligence theory. Accordingly, the former Fifth Circuit rejected such a claim in a strict liability action. Challoner v. Day & Zimmermann, Inc., 512 F.2d 77 (5th Cir.1975), vacated on other grounds, 423 U.S. 3, 96 S. Ct. 167, 46 L. Ed. 2d 3 (1975), remanded, 546 F.2d 26 (5th Cir.1977).4

The contract specifications "defense" is not asserted in this case, so we do not consider its application. In suits involving military suppliers who participate in--and thus have knowledge of--product design, as in this action, it is in any event unlikely to be available.

A second and analytically distinct defense is the "government agency defense." It grows out of the Supreme Court's decision in Yearsley v. W.A. Ross Construction Co., 309 U.S. 18, 60 S. Ct. 413, 84 L. Ed. 554 (1940), in which the Court absolved from liability a contractor who, at the request of the government, built dikes in the Missouri River and accidentally washed away part of petitioners' land. The Yearsley court apparently regarded this contractor as "an agent or officer of the government," acting on the government's behalf. Id. at 21, 60 S.Ct. at 414. Since "[t]he action of the agent is 'the act of the government,' " id. at 22, 60 S.Ct. at 415, the contractor could be deemed to share in federal sovereign immunity. Although such immunity has been waived in many cases, where injuries to military personnel incident to service result from defective product design, the government may not be sued for damages under the Feres doctrine. See Feres v. United States, 340 U.S. 135, 71 S. Ct. 153, 95 L. Ed. 152 (1950); Stencel Aero Engineering Corp. v. United States, 431 U.S. 666, 97 S. Ct. 2054, 52 L. Ed. 2d 665 (1977).5

This defense is rarely invoked, and its elements are nowhere clearly stated. To evaluate a Yearsley claim in the military contractor context, a court would appear to be obliged to take three steps. First, it would apply the Feres doctrine to determine whether the government itself could be sued in that situation. If not, the court would then invoke a second body of doctrine, the law of principal and agent, to inquire whether the contractor actually acted as an agent of the government. Precedent in this Court makes it clear that, at least in manufacturing defect cases arising in a military setting, a firm must be more than simply an independent contractor to be regarded as the government's agent. In Whitaker v. Harvell-Kilgore Corp., 418 F.2d 1010, 1013-1015 (5th Cir.1969), the former Fifth Circuit rejected the sovereign immunity claim of such a contractor that manufactured a defective grenade. Finally, if the contractor proved agency status, the court would inquire whether the agent acted within the course and scope of its duties.

We decline appellant's invitation to consider its defense under Yearsley, though we note that the Whitaker reasoning would in any case logically control in design as well as manufacturing defect cases. The agency/sovereign immunity theory was neither presented fully to nor ruled on by the court below.

We now consider the "military contractor defense" asserted in this case. As that defense is recognized in this Circuit, it is neither the same as, nor does it grow out of, either of the other two theories--contrary to the conviction of many. See Bynum, at 561 (the government contractor defense "is essentially an amalgamation of the two traditional defenses [contract specifications and Yearsley ]"); McKay, 704 F.2d at 448 ("[t]his rule was first articulated by the Supreme Court in Yearsley...."). The military contractor defense is available in certain situations not because a contractor is appropriately held to a reduced standard of care, nor because it is cloaked with sovereign immunity, but because traditional separation of powers doctrine compels the defense. Judge Pratt put this point succinctly in In re "Agent Orange" Product Liability Litigation, 534 F. Supp. 1046, 1054 n. 1 (E.D.N.Y.1982):

The purpose of a government contract defense ... is to permit the government to wage war in whatever manner the government deems advisable, and to do so with the support of suppliers of military weapons. Considerations of cost, time of production, risks to participants, risks to third parties, and any other factors that might weigh on the decisions of whether, when, and how to use a particular weapon, are uniquely questions for the military and are exempt from review by civilian courts.

We add that the purpose of such a defense is not only to permit the government actually to "wage war," but also to prepare for war--that is, this country need not be in combat for the defense to obtain. In making this determination, we do not implicitly sanction any and every military judgment involving weapons or weapons systems6 potentially dangerous to servicemen. We simply acknowledge that in our constitutional system, when a knowing and purposeful decision to employ such products is made by the military, the judiciary may not question it.7 The public, of course, retains its right to dispute or to endorse those decisions--from the use of "Agent Orange" to the military's choice of aircraft--in the court of public opinion or at the polls.

In acknowledging our obligations under separation of powers doctrine, we are also mindful of the longstanding commitment of our judicial system--developed in the common law of torts, and more recently in interpretation of strict product liability statutes--to the fair compensation of the individual victim of defectively designed products.8 Thus, in recognizing a defense to liability for military contractors, we carefully frame a narrow exception to the product liability law that governs all other product designers. In so doing, we are compelled to examine rigorously, and at some points to reject, the rationales behind the defense and the elements necessary to establish it as prescribed by other circuits.

The leading version of the military contractor defense was set out by the Ninth Circuit in McKay, 704 F.2d at 448-451.9 The McKay court declared that "[t]he reasons for applying the government contractor defense to suppliers of military equipment with design defects approved by the government parallel those supporting the Feres-Stencel doctrine." Id. at 449. Although in our view the military contract defense may be applied quite apart from Feres --that is to say, its separation of powers rationale is independently coherent--we accept the parallel. However, since the Feres doctrine has undergone a certain amount of reworking since its inception and, indeed, even since McKay, it is necessary to reframe the McKay rationales somewhat.

The McKay court articulated four such policy rationales to support the military contractor defense based on Feres. First, it claimed that holding a military supplier liable in these cases would "subvert the Feres-Stencel rule," because those contractors would pass on the cost of accidents to the government in their contract prices. Id. at 449. Second, the McKay court reasoned that to hold military suppliers liable for defective designs where the government set or approved the design specifications would improperly "thrust the judiciary into the making of military decisions." Id. Third, the court concluded that imposing liability on the contractor would undermine the nation's "ability to push technology to its limits and thereby to incur risks beyond those that would be acceptable for ordinary consumer goods." Id. at 450. Finally, the court suggested that the defense encourages suppliers to work closely with military authorities in the development and testing of equipment, so that responsibility for product design is more easily fixed. Id.

We regard the first McKay rationale as a weak support for the government contractor defense for several reasons. First, we are not convinced that the cost pass-through rationale is economically sound. To the extent that any competition obtains in the market for defense products, for example, contractors with defective designs may be deterred from passing through the cost of liability10 for defective design11 by competition from contractors with better safety records. See McKay, 704 F.2d at 457 (Alarcon, J., dissenting). And even if there is some transfer of costs, the net price to the government of contractor liability in these cases might be less without a tort law exception--that is, if legal incentives promote better-designed planes and fewer costly accidents. (The loss from the crash of a single A-6 is in the millions of dollars--not to mention the cost in human terms.)

Second, economically sound or not, the cost pass-through rationale is based on a strained reading of Stencel and an outdated interpretation of Feres. What the Court refused to let in the "back door" in Stencel, 431 U.S. at 673, 97 S.Ct. at 2058, in which the Court barred a cross-claim against the government by a military contractor sued for product liability, was full and compulsory indemnification for the serviceman's claim. More adventitious pass-through of some portion of the cost of product liability insurance by a contractor presents a considerably attenuated problem. Stencel itself at least arguably sanctioned some cost pass-through by contractors. See id. at 674 n. 8, 97 S. Ct. at 2059 n. 8 (the contractor "no doubt had sufficient notice so as to take this risk into account in negotiating its contract for the emergency eject system at issue here"). See also McKay, 704 F.2d at 457 (Alarcon, J., dissenting).

But in any case, the limitation of government liability rationale behind the Feres-Stencel doctrine appears, since a recent Supreme Court decision, to be "no longer controlling." United States v. Shearer, --- U.S. ----, 105 S. Ct. 3039, 3043 n. 4, 87 L. Ed. 2d 38 (1985). Thus, this Circuit will not recognize a military contractor defense simply to shield such contractors--and (only arguably) the government--from the cost of liability for defectively designed products.

As we noted above, there are better reasons for the defense. The factors controlling in Feres inquiries that do survive Shearer are (1) "whether the suit requires the court to second-guess military decisions," and (2) "whether the suit might impair essential military discipline." Id. 105 S.Ct. at 3043. These concerns are often conflated into a single "military discipline rationale" for the Feres doctrine. But the two strands of analysis are distinct. The former is the classic separation of powers theory described above. Applied to the military supplier context, it means simply that the military must be free to decide that the risk of accident or injury to servicemen from a design defect in the equipment it orders is a risk that it is willing to take--without judicial interference. This is the basis for the military contractor defense in this Circuit.

The latter "military discipline" strand was most clearly set out in United States v. Brown, 348 U.S. 110, 112, 75 S. Ct. 141, 143, 99 L. Ed. 139 (1954), in which the Court said that "[t]he peculiar relationship of the soldier to his superiors, the effects of the maintenance of such suits on discipline, and the extreme results that might obtain if suits under the Tort Claims Act were allowed for negligent orders given or negligent acts committed in the course of military duty" underlies Feres ' grant of immunity. Two concerns are actually involved here: (1) the notion that a soldier might use the civilian courts to challenge the act or order of a superior officer; and (2) the idea that in a civilian suit of any sort involving a serviceman, members of the military might be compelled to testify against one another, to the detriment of discipline. The former quite simply would not justify a military contractor defense since the challenge in these cases is to an independent contractor, not a military officer.12 (Insofar as a military "order" is implicated--that is, the original decision to buy and use the product at issue--it is covered under the military decision strand, above.)

The latter concern is also not present in military contractor suits in quite the same way as it is under Feres. The Feres exception to the FTCA renders the United States government immune from suit, so that potentially disruptive testimony by fellow servicemen is for the most part simply precluded. The military contractor defense is, in contrast, an affirmative defense that must be pled and proved by defendant in the course of suit. In other words, whether the defense is asserted or not, the plaintiff will in any event put on his or her case. Indeed, it may be the raising of the defense, rather than the maintenance of the action itself, that makes potentially conflicting military testimony necessary--a point which, of course, argues in favor of liability for contractors, and against the availability of the military contractor defense. However, since we think the likelihood of any profound disruption of discipline is negligible from testimony in suits against military contractors, we decline to rest either recognition or rejection of the defense on this frail support. Thus, we find, as this Court did in Cole v. United States, 755 F.2d 873, 879 (11th Cir.1985) (need to protect military discipline does not extend to suits involving veterans), that the danger of interfering with discipline in military contractor cases "is too remote to be accorded significant weight when the decision only indirectly involves military orders or practices concerning active duty soldiers."

We turn finally to the last two rationales used by the McKay court to justify the military contractor defense. That court's concern that the imposition of liability on contractors would chill risk-taking in the development of new technology is, in our view, no more than a restatement of the military decision rationale outlined above. Under separation of powers theory, military risk-taking--where it involves products supplied by contractors--is shielded from judicial scrutiny by the military contractor defense, provided only that it is knowing and purposeful. To adopt the new technology argument as an independent rationale for the defense is not only to be redundant, but also to invite unproductive quibbling over whether or not a particular product is actually on the cutting edge.

We find the last McKay rationale--the notion that the military contractor defense encourages the military and its suppliers to work closely together, thereby making it easier to discover who is responsible for product design--somewhat inscrutable. Indeed, on the contrary, our experience is that the more closely the contractor and the military work together, the more difficult it is to determine exactly who made design decisions. In any case, as it is now limited by Shearer, Feres provides little support for this rationale.

In sum, we recognize a military contractor defense in this Circuit based exclusively on the theory that the constitutional separation of powers compels the judiciary to defer to a military decision to use a weapon or weapons system (or a part thereof) designed by an independent contractor, despite its risks to servicemen.13 The court's task is to inquire whether such a decision has been made and, if not, to apply the standard principles of product liability law.

Courts have adopted two basic forms of the government contractor defense, and one hybrid version. The leading standard, again, was set out by the McKay court. McKay, 704 F.2d at 451. That rule is as follows:

[A] supplier of military equipment is not subject to section 402A [strict] liability for a design defect where: (1) the United States is immune from liability under Feres and Stencel, (2) the supplier proves that the United States established, or approved, reasonably precise specifications for the allegedly defective military equipment, (3) the equipment conformed to those specifications, and (4) the supplier warned the United States about patent errors in the government's specifications or about the dangers involved in the use of the equipment that were known to the supplier but not to the United States.

In Tillett and Bynum, the Seventh and Fifth Circuits adopted the McKay test without modification. This is the test appellant advocates and the one, in fact, that the court below applied (although perhaps incorrectly).14

The other major version of the standard was announced in In Re "Agent Orange" Product Liability Litigation, 534 F. Supp. at 1055 (manufacturers of chemical defoliant not liable to American servicemen for effects of exposure). The three elements of that defense are: (1) proof that the government established contract specifications for the product; (2) proof that the contractor complied in all material respects with the specifications; and (3) proof that the government's knowledge of the hazards of the finished product was at least equal to that of the contractor. Id. As part of the third element, the court imposed a duty on the manufacturer to warn of the product's dangers. Id. at 1058.

The critical difference between the McKay and the Agent Orange tests is the degree to which the government may be involved in setting the product specifications. McKay seems to acknowledge that there is some de minimis level of government involvement below which the supplier should be held responsible for design defects. McKay, 704 F.2d at 450 ("[w]hen only minimal or very general requirements are set for the contractor by the United States the rule is inapplicable"). However, under the McKay test mere government approval of a contractor's design will shield the supplier from liability. No such provision obtains in the original Agent Orange test.

In Koutsoubos the Third Circuit introduced a kind of hybrid standard. It formally adopted the Agent Orange test, above, 755 F.2d at 355, but later read into its first part the McKay "approval" reasoning--that is, as the court said, "government approval of specifications developed through a continuous series of negotiations between the contractor and the military [would] satisf[y] the first prong of the Agent Orange defense, even though the majority of specifications originated with the contractor," In Re: Air Crash Disaster, 769 F.2d at 122. Thus, this test is virtually the same as the McKay defense.

We are not satisfied with either of these two standards. Our central concern is whether or not the military actually made a decision to use a product that it knew to be dangerous to servicemen. As the Third Circuit admitted in In Re: Air Crash Disaster, id. at 121, "no separation of powers concern arises unless the design in question represents a judgment by the military." Since the military and its representatives are not before us in these suits against contractors, we may not ask them for proof that such a decision was actually made. We may only require from the contractor proof of the negative proposition--that the design in question does not merely represent a judgment by the supplier.

In fashioning a standard by which courts may measure responsibility for design judgments, we consciously avoid too sanguine a view of the usefulness of product "specifications." Specifications may be minimal or detailed, quantitative or qualitative, general or specific; they may range from meticulous descriptions of each bearing and bushing required, to vague hopes for "simple" or "failsafe" products. At times, several sets of specifications, sometimes conflicting, may govern a product's design all at once: e.g., one "spec" requiring back-up or redundancy systems in all products, another urging ease of maintenance, a third mandating combat effectiveness, a fourth seeking cost containment, and a fifth prescribing the dimensions of a washer. Worse still, these specifications may be promulgated by several different sources, military or civilian, at different times over the life of a product.

Although we are conscious of some overlap, for purposes of our analysis we divide specifications into two types: (1) detailed, precise and typically quantitative15 specifications for manufacture of a particular military product--that is to say, the design or blueprint for production; and (2) more general and more qualitative specifications, such as performance or mission criteria--that is to say, all other specs besides type one. If the type one design is generated exclusively by the military, then the product supplier is in the position of a manufacturer rather than a designer and, thus, is liable only for manufacturing defects (that is, failure to meet these detailed specifications).

In every other instance in which a contractor participates in the creation of a military product, it must be, in some respect, itself a producer of the type one product design--that is, one responding to general, qualitative military specifications with more precise and quantitative design specifications. Where the resulting product "fails" (crashes, explodes, aborts or otherwise injures someone) the design a priori is unlikely to meet the military's general, qualitative specifications: that is, it has not in fact accomplished its mission, or performed properly, simply, or safely.16 Thus, the central inquiry of the McKay and Agent Orange tests--does the design conform to reasonably precise military specifications?--is, depending on the sort of specification we mean, either a question that is simply irrelevant to a design defect case (since an affirmative answer makes the contractor no more than a manufacturer), or a question begged by the very occurrence of the accident itself.

Thus, we fashion a test with a somewhat different focus. As a general rule, the military contractor will be liable, as is any other contractor, to servicemen injured by defects (i.e., unreasonable dangerousness) in the products, or portions or phases of products, that it designs--that is, for which it provides type one specifications. A contractor may escape liability only if it affirmatively proves: (1) that it did not participate, or participated only minimally, in the design of those products or parts of products shown to be defective; or (2) that it timely warned the military of the risks of the design and notified it of alternative designs reasonably known by the contractor, and that the military, although forewarned, clearly authorized the contractor to proceed with the dangerous design.

Element one of this test is fashioned to allow a military contractor to show, when it actually works jointly with military personnel in producing type one product specifications, that the part it played was so minimal as to excuse it from proving the second part of the test. This determination is appropriately made by the trial judge. To structure the test in this way is self-consciously to turn the assumption behind the McKay test on its head. Here, we do not ask the contractor to prove that the government prepared the specs and find that the contractor's participation in their development through "continuous back-and-forth" will not defeat the defense, Koutsoubos, 755 F.2d at 355. Rather, we require the contractor to show that it did not prepare the specs, and hold that the government's participation, if sufficiently great, may prove the defense.

If element one of this defense is not proved, as in any other affirmative defense, by a preponderance of the evidence, then element two first requires a demonstration that the contractor both warned the military of the risks of the product or product part that it designed and informed it of design alternatives "reasonably known" to the contractor. Reasonable knowledge is not synonymous with omniscience. A risk is reasonably known when it is either actually known, or reasonably ought to be known given good design practice in the industry.17 Similarly, an alternative is reasonably known if it is either actually known, or reasonably ought to be known given good design practice in the industry. Whether the contractor's counsel on these matters is sufficiently specific and complete to permit an informed decision on the part of the military is a matter of fact for the trial judge to determine.18 In making this assessment of sufficiency, the court may take into account evidence that goes to the military's own level of relevant knowledge and expertise.

Element two further requires a showing by the contractor "that the military, although forewarned, clearly authorized the contractor to proceed with the dangerous design." Authorization here must be knowing--that is, more than "rubber stamp" approval. It must also be clear--that is, obviously related and responsive to the relevant warning.

The over-riding objective of this test, again, is to determine whether or not a military judgment to go ahead with a dangerous design was actually made. If so, the contractor that created or helped create the design is absolved from judicially-imposed liability. If not, then the contractor is subject to the customary strictures of product liability law.

Two design defects are alleged in the case at hand: (1) Grumman's failure to provide a back-up or warning system in the event of a "disconnect" in the longitudinal flight control system,19 and (2) Grumman's failure to correct the disconnect problem with its proposal for "self-retaining" bolts. Grumman does not prove the elements of the military contractor defense as to either defect.

Element one requires proof that Grumman did not participate, or participated only minimally, in the design (that is, the production of type one specifications) of the defective product or part. But the trial court found that Grumman exclusively designed and produced detailed specifications for this aircraft. Grumman also suggested, in engineering change proposals (ECP's), the use of self-retaining bolts. The trial court's finding that Grumman was responsible for the A-6's design and alteration is not clearly erroneous.

Element two provides that a military contractor may still defend itself against liability if it timely informed the military of the design risks and alternatives reasonably known to it, and if the military, so warned, clearly authorized the contractor to proceed. The trial court found that Grumman knew of the defect in the longitudinal control system of the aircraft. It found, further, that Grumman failed to warn the Navy sufficiently of the danger inherent in the non-redundant longitudinal control system. The lower court noted that Grumman was both actually aware and reasonably should have been aware, given good design practice, of the risk of non-redundancy. And it found that the Navy did not have sufficient expertise to render a warning of that risk unnecessary.

As to the self-retaining bolts, the trial court also found that the Navy relied on Grumman's advice that these would solve the problem. This may be read as a finding of a failure to warn that the bolts would not correct the underlying defect. None of these findings is clearly erroneous.

Thus, although the Navy did formally approve Grumman's A-6 specifications and design changes, that approval did not constitute the sort of informed military decision to accept the risk of a dangerous product to which this Court must defer under separation of powers doctrine.

Appellee therefore recovers, and the judgment of the lower court is AFFIRMED.