4
Patents, Trade Secrets, Digital Products, and Federal Laboratories
As discussed in Chapter 3, digital products created by federal laboratories may be eligible for different forms of intellectual property rights (IPRs). Article I, Section 8, Clause 8 of the U.S. Constitution authorizes Congress to enact patent and copyright laws “To promote the progress of science and useful arts”—a clear instrumentalist justification. It empowers Congress to grant limited periods of exclusivity to “authors and inventors” to promote “progress,” understood to include the creation of new works. This chapter examines federal labs’ use of patents and trade secrets for digital products and the impact on the dissemination and commercialization of those products. Noted in the discussion is that many valuable digital products created by federal labs lack formal IP protection and that labs may compensate for this lack of legal protection by imposing restrictive contractual terms and technological control mechanisms on those products.
PATENTS
Historically, patents have been a primary tool utilized by federal laboratories to protect inventions and license rights to those inventions to the private sector. Beginning with a historical overview of the use of patents for digital products, this section describes federal practices and policies with respect to such products developed with the involvement of federal labs and offers a brief discussion of federal labs’ reporting on patents.
History of Patents on Digital Products
A primary goal of patent law is to encourage and enhance innovation and the commercialization of new technologies. Patent protection is possible only when an inventor applies for and is granted a patent by the U.S. Patent and Trademark Office (USPTO), a process that tends to be expensive, complex,
difficult, and time-consuming. In most cases, digital content, audiovisual works, data, and the like do not qualify as patentable subject matter, but there are exceptions. This section looks at two types of digital products that have generated significant controversy with respect to patenting: software and algorithms, and scientific and technical data.
Software and Algorithms
The patent eligibility of computer software and algorithms has fluctuated over time. It has long been the case that abstract ideas, such as mathematical formulae, are not eligible as patent subject matter. In Gottschalk v. Benson, 403 U.S. 63 (1972), the Supreme Court rejected a patent claiming “a method for converting binary-coded-decimal…numerals into pure binary numerals” using a general-purpose digital computer. The Court reasoned that the “claim is so abstract and sweeping as to cover both known and unknown uses of the…conversion [method].” As a result, the claim was considered to consist of abstract ideas that were ineligible for patent protection. Six years later, in Parker v. Flook, 437 U.S. 584 (1978), the Court held that a patent claiming several conventional applications of a novel mathematical formula was similarly drawn to ineligible subject matter.
It was not until 1981, in Diamond v. Diehr, 450 U.S. 175 (1981), that the Supreme Court upheld a patent claiming computer software. The claimed method employed the well-known Arrhenius equation to calculate and control the temperature in a process for curing rubber. The Court held that while the Arrhenius equation itself was not patentable, the claimed method for curing rubber was an industrial process of a type that has historically enjoyed patent protection. The use of the equation and a computer were incidental to the patentable inventive process.
Although the requirements for patenting software fluctuated over the next three decades, large numbers of such patents were issued. Software patents differ substantially from copyrights covering computer software. Copyright protects the expression of a work—the lines of code written by a programmer, the executable version of that code, and the screen displays and images generated by the code. Patents, on the other hand, protect software functionality at a higher level. Actual source code is seldom included in a patent application. In principle, the disclosure requirements of patent law counsel that flowcharts showing the logical structure of a program should be included (Rai, 2012). Given the severe time limitations faced by patent examiners, however, their enforcement of patentability requirements is often subpar (Frakes and Wasserman, 2017), and courts have not enforced the principle that software disclosure requires something more than high-level, generalized information about the functions accomplished (but see Allison and Ouellette, 2016). As a consequence, software patents in many cases simply describe and claim the functions accomplished by particular programs (Lemley, 2013).
The vagueness, potential overbreadth, and overall poor quality of many software patents led to significant criticism of software patenting in the 2000s. Notorious examples of questionable software patents emerged, including Amazon’s “one-click shopping” patent, British Telecom’s patent that allegedly covered “the Internet,” and Apple’s patents covering such basic smartphone gestures as “tap to zoom.” Compounding these issues, the 2000s also saw the rise of significant patent litigation initiated by so-called patent assertion (or nonpracticing) entities (known colloquially as “patent trolls”) that took advantage of overly broad and vague patent claim language to seek monetary settlements from many firms in the electronics and computing industry. The patent system came under heavy fire from the popular media, scholars, and even the Obama Administration (Burk and Lemley, 2009; Bessen and Meurer, 2009; Rai, 2012, pp. 1274–1277; Executive Office of the President, 2013).
Perhaps in response to some of these issues, the Supreme Court again turned its attention to algorithmic patents in 2010. In Bilski v. Kappos, 561 U.S. 593 (2010), the Court invoked the requirement of patentable subject matter to hold that an algorithm for calculating a fixed price for monthly utility bills was an unpatentable abstract idea. Then, in Alice Corp. v. CLS Bank International, 573 U.S. 208 (2014), the Court rejected “patent claims drawn to a computer-implemented, electronic escrow service for facilitating financial transactions,” holding that the invention was merely an abstract idea. The Court also observed that merely including a generic computer implementation of such an abstract idea fails to transform it into a patent-eligible invention.
The holding in Alice overturned much existing wisdom and practice regarding software patenting. The holding appears to have led, at least initially, to a steep increase in the number of cases in which patent or patent application claims for software and related digital products, particularly in the area of software covering business methods, were found to be invalid (Tran, 2016; Chien and Wu, 2018).1 Many in the patent bar and judiciary have argued that the standard that was set in Alice is overly ambiguous and indeterminate (Reinecke, 2019).
Because of definitional challenges, rigorous empirical assessment of the longer-term impact of Alice on software patenting is difficult. Over the course of the decades-long controversy surrounding software patents detailed above, researchers attempting to examine this question have used differing definitions for software, and their results have differed substantially based on the definition used (Rai et al., 2009). In 2013, the Office of the Chief Economist at the USPTO, working with patent examiners, developed a definition of software based on USPTO classifications, which is used in this report (Graham and Vishnubhakat, 2013). But this definition is far from definitive. Because software is a “general-purpose” technology, now utilized in many if not most areas of invention, distinguishing “pure” software capable of performing on many platforms from
___________________
1 Also contributing to this increase was the introduction in 2012 of a new procedure for challenging patents at the Patent Trial and Appeals Board (PTAB), known as inter partes review.
software inextricably linked to a hardware product can be difficult (Rai et al., 2009).
Categorization issues aside, one industry source that purports to utilize the 2013 Graham/Vishnubhakat definition reports that by 2019, more than 60 percent of all newly issued U.S. patents covered software-related inventions (Millien, 2020). In addition, perhaps because of USPTO guidance and Federal Circuit decisions that interpret Alice narrowly, the number of successful challenges to software claims appears to have declined in the years immediately after the Alice decision was rendered (Saltiel, 2019; Tran and Benevento, 2019). Thus, patenting activity in the sector continues.
Scientific and Technical Data
The federal government has made, and continues to make, large quantities of scientific data available to the public. Typical government-generated data, such as those describing astronomical, geological, and meteorological observations, are not considered to be useful inventions or discoveries that are eligible for patent protection.2 In some cases, however, scientific and technical data can be closely linked to physical materials or processes and may thus be claimed as part of a patented invention.3
Attempts to include scientific data in patent claims have probably attracted the greatest attention in the area of human genomics. DNA is an organic molecule comprising two intertwined chains of four nucleotide bases (cytosine [C], guanine [G], adenine [A], and thymine [T]), a sugar called deoxyribose, and a phosphate group. One of DNA’s principal functions within cells is to code for proteins, a function that is dictated by the precise order in which the nucleotide bases occur within segments of the DNA molecule known as genes. Thus, while genes and other DNA segments are organic molecules, the information conveyed by their nucleotide sequences has significant importance and commercial value.
Accordingly, by the late 1980s, researchers began to seek patents claiming DNA sequence information (Eisenberg, 1990; NRC, 2006). As technological capacity to sequence DNA increased, researchers appended voluminous sequence listings, often running more than 100 printed pages, to their patent applications. The growth of patents on human DNA sequences, often referred to as “gene patents,” caused concern in some quarters, and by 2005, two Massachusetts Institute of Technology (MIT) researchers estimated that a full 20 percent of human genes were covered by patent claims to some degree (Jensen and Murray, 2005).
During this period, two significant legal developments limited the ability of researchers to patent DNA sequence information. The first arose under the
___________________
2 In this sense, data should be distinguished from data structures, which have been recognized as eligible patent subject matter, at least prior to Alice (Hollander, 2003).
3 The USPTO specifically provides instructions for including data tables, even those that may exceed 100 pages in length, in patent applications (MPEP § 608.05).
“utility” requirement of Section 101 of the Patent Act, which mandates that to be patentable, an invention must be “useful.” As the rate of DNA sequences “discovered” in the laboratory increased, patent claims on such discoveries also increased. Yet in many cases, the precise biological function of these sequences was unknown, so it was not possible to describe their utility. Patent attorneys got around this obstacle by arguing that these sequences could be useful in further research as markers or locators for particular DNA segments. But by 1999, in response to significant criticism of such patents, the USPTO released draft Utility Guidelines that required inventions to have “specific, substantial, and credible” utility in order to be patented. This requirement became official USPTO policy in 2001, and was confirmed a few years later by the Federal Circuit in In re Fisher, 421 F.3d 1365 (2005).
The second major blow to DNA patenting was dealt by the Supreme Court in Association for Molecular Pathology v. Myriad Genetics, Inc., 569 U.S. 576 (2013). In that closely watched case, the Court held that the nucleotide sequences of the isolated and purified BRCA1/2 human genes were not patent-eligible subject matter, as they constituted naturally occurring substances found in every human cell. Thus, despite the patentee’s insistence that its claims covered an organic molecule that in its isolated and purified form is found nowhere in nature, the Court reasoned that “the [patent] claims are not expressed in terms of chemical composition, nor do they rely on the chemical changes resulting from the isolation of a particular DNA section. Instead, they focus on the genetic information encoded in the BRCA1 and BRCA2 genes” (emphasis added). And in the Court’s view, that information was not eligible for patent protection. Accordingly, DNA sequence data today are largely considered ineligible for patent protection.
Patents and Federal Digital Products
Federal laboratories have patented digital products for many years. This section reviews federal patenting practices and policies, first generally and then in connection with genomic data and software. It addresses both federally owned inventions and inventions owned by third parties but generated with contributions from federal researchers.
Patents Claiming Federal Inventions
In the United States, only an inventor(s) may apply for a patent on a particular invention. Typically, if an inventor is employed and the invention is within the scope of his or her employment, the inventor will assign ownership of the resulting patent to his or her employer. In contrast with the work-made-for-hire doctrine under copyright law, however, an employer does not automatically obtain ownership rights in inventions created by its employees.
Federal agencies may obtain ownership rights in inventions either through contractual assignments from their employees or contractors or by
operation of law when a government-funded institution fails to make a disclosure or election to retain rights in an invention (35 U.S.C. § 202[c]).4 Federal agencies are permitted to apply for, obtain, and maintain patents in the United States and foreign countries that claim inventions they own (35 U.S.C. § 207[a][1]). See Box 4-1 for more information on Section 209 of the Bayh-Dole Act.
Patents that arise from cooperative research and development agreements (CRADAs) (35 U.S.C. § 3710[a] et seq.) are another important category of patents closely tied to intramural research. Under the 1986 amendments to the Stevenson-Wydler Act, a federal laboratory that is a party to a CRADA may grant to its collaborating party an exclusive license or assignment for any inventions created by employees of the lab. However, these patents do not appear to be tracked under any existing reporting mechanism (see Chapter 7 for further discussion).
Genomic Data and Federal Patent Deterrence Strategies
Starting in 1991, researchers at the National Institutes of Health (NIH) began to file patent applications covering short DNA segments known as “expressed sequence tags” (ESTs), arguing that they could help locate much larger genes more quickly and efficiently than would sequencing the entire genome, making NIH one of the first agencies to seek patents on human DNA sequences. After some initial controversy, the agency abandoned its patenting efforts with respect to ESTs (IOM, 2003; Rai, 2012) and actively campaigned to have the USPTO amend its Utility Guidelines to disallow the patenting of DNA
___________________
4 The provisions in question form part of the Bayh-Dole Act of 1980, Pub. L. No. 96-517, which is generally discussed in the context of patenting by federally funded universities and other institutions. However, §§ 207–209 of the act govern the patenting and licensing of inventions by federal agencies.
sequences of unknown function. This position eventually emerged in the USPTO’s 1999 draft guidelines (NRC, 2006; Rai, 2012).
NIH opposition to DNA-based patents next manifested in the data release policies adopted during the Human Genome Project (HGP)—specifically, the adoption of the 1996 Bermuda Principles on data sharing as official agency policy. This policy required that genomic sequence data be released into NIH’s GenBank or another public database within 24 hours of being generated, a significant commitment to public data sharing. This rapid, public release of HGP data not only represented the agency’s disavowal of patenting but also was intended to deter third-party patenting by making the sequences prior art (Contreras, 2011).
In its 1996 policy adopting the Bermuda Principles, moreover, NIH warned that the agency “will monitor grantee activity…to learn whether or not attempts are being made to patent large blocks of primary human genomic DNA sequence” (NHGRI, 1996). And in its 2005 Best Practices for the Licensing of Genomic Inventions, the agency advised grantees to be cautious in seeking patents for DNA-based inventions (NIH, 2005). However, it is not clear that NIH has ever actively monitored or enforced compliance with these policies (Contreras, 2017; Rai and Eisenberg, 2003). The Bayh-Dole Act specifies a set of procedures for constraining patenting by grantees, and the NIH policies may accordingly have hortatory rather than legal significance.
Following the completion of the HGP in 2003, NIH continued to fund and curate ever-growing collections of genomic data (Contreras and Knoppers, 2018). As the types of data included in public databases grew more complex and began to encompass clinical, phenotypic, and demographic data, data access policies and procedures expanded to accommodate this new landscape. Through about 2008, NIH continued to insist upon rapid data release mechanisms that would limit the patentability of research data and discoveries. In more recent policies, however, the agency’s advocacy for patent deterrence appears to have weakened, perhaps because since Myriad, patents no longer have posed a significant threat to the use of genomic data in research (Contreras and Knoppers, 2018). Further, despite NIH’s overt opposition to patents covering raw DNA sequence data and its public release of large quantities of genomic data, the federal government remains a major holder of patents covering more specifically focused DNA-based inventions. According to one study, by 2005 the federal government held more DNA-based patents than any entity other than the University of California system (NRC, 2006), and as of 2009 it remained in fifth place (Cook-Deegan and Heaney, 2010).
Patenting of Software by Federal Laboratories
As discussed in Chapter 7, agency reporting of federal laboratory patents to the National Institute of Standards and Technology (NIST) does not identify the lab or whether the patent is for a digital product. Moreover, as noted above,
defining what constitutes a software patent is difficult. Nonetheless, the committee attempted to gather as much data on this issue as possible.
Specifically, using the definition articulated by Graham and Vishnubhakat (2013), the committee examined software patenting between 1980 and 2014 by the Department of Energy (DOE), the Department of Defense (DOD), and the National Aeronautics and Space Administration (NASA). Starting in the 1990s, these three agencies began to secure a total of at least several hundred patents annually. Given that federal labs as a whole secure only a few thousand patents annually across all technologies, it appears that software patenting makes up a nontrivial percentage of the patent activity of federal labs.
That said, the committee’s data encompass activity only through 2014, when the Alice case was decided by the Supreme Court. Accordingly, these data are consistent with the possibility, advanced by some federal lab representatives, that few labs currently seek software patents. Additionally, even the number of software patents sought and secured by federal labs pre-Alice may represent only a small percentage of the total software produced by the labs. As noted, the government releases a significant amount of software on an open-source basis.
Licensing of Inventions Generated with Federal Investment
Federal agencies are authorized to “grant nonexclusive, exclusive, or partially exclusive licenses under federally owned inventions, royalty-free or for royalties or other consideration, and on such terms and conditions…as determined appropriate in the public interest” (35 U.S.C. § 207[a][2]). Because the development of federal inventions was funded by U.S. taxpayers, such licenses are intended to benefit the U.S. economy.
Nonexclusive licenses, under which a patent holder grants the right to operate under a patent to multiple users, are generally viewed as favorable to broad dissemination of a given technology. NIH, for example, advises that “whenever possible, non-exclusive licensing should be pursued as a best practice” (NIH, 2005). An exclusive license, however, can provide the licensee with an incentive to make a significant financial investment in commercializing the licensed technology. As NIH notes, “in those cases where exclusive licensing is necessary to encourage research and development by private partners, best practices dictate that exclusive licenses should be appropriately tailored to ensure expeditious development of as many aspects of the technology as possible” (NIH, 2005). Contractual mechanisms for ensuring expeditious development include, among others, the imposition on the exclusive licensee of “diligence” milestones—commercial or technological goals that must be met within specified time periods, or the exclusivity or the license itself may be terminated. To this end, 35 U.S.C. § 209(b) establishes an expectation (though not a firm requirement) that exclusive licenses of federally developed technology shall “normally” be granted “only to a licensee who agrees that any products
embodying the invention or produced through the use of the invention will be manufactured substantially in the United States.”
If a federal agency wishes to grant a license that is exclusive or partially exclusive, 35 U.S.C. § 209(a) imposes a number of additional requirements on the agency; however, these requirements apply only to government-owned, government-operated (GOGO) labs. Specifically, (1) the exclusive license must be a “reasonable and necessary incentive to call forth the investment capital and expenditures needed to bring the invention to practical application or otherwise promote the invention’s utilization by the public”; (2) the public must be served by the granting of the license, “as indicated by the applicant’s intentions, plans, and ability to bring the invention to practical application or otherwise promote the invention’s utilization by the public”; and (3) the agency must ensure that the scope of exclusivity is no greater than reasonably necessary to achieve these goals. Similarly, the exclusive licensee must commit “to achieve practical application of the invention within a reasonable time.” Finally, the agency must ensure that “granting the license will not tend to substantially lessen competition or create or maintain a violation of the Federal antitrust laws.” The agency is authorized to terminate the license if any of these provisions is violated.5
In sum, these provisions require each federal agency to ensure that the licensing of its inventions and related patents is accomplished in a manner that supports the public interest, recognizing the general principles discussed in the introduction to this chapter. Exclusive licenses are subject to particular scrutiny as they allow an agency to confer the benefits of a federal invention—developed at taxpayer expense—on a single licensee to the exclusion of all others. Thus, the agency must determine that the limitations on broad adoption of the invention that arise from an exclusive license are justified and serve the public interest, rather than merely the commercial interests of the licensee and the agency.
Some federal agencies have taken the position that granting an exclusive license to a fundamental technology can confer a substantial commercial advantage on a single licensee at the expense of the overall market and the advancement of science and technology. However, a recent article identifies cases in which federal agencies have granted exclusive licenses on federal inventions to patent assertion entities (PAEs) (Reilly and Waxman, 2016). This article reports that between 2006 and 2015, four federal agencies (the National Security Agency [NSA], NASA, the U.S. Navy, and the U.S. Air Force) granted exclusive or partially exclusive licenses covering more than 200 federal patents to seven different PAEs. As has been widely reported in the popular media, PAEs are extremely controversial to the extent that they extract monetary licensing fees or litigation settlements from existing market actors, and do not themselves innovate, utilize technology, or bring products to market. Reilly and Waxman (2016) report that during 2006–2015, in exchange for exclusive rights to federal inventions, the agencies received from the PAEs up-front payments, including at least one in excess of $2 million, as well as ongoing royalties. Further, the licensing activities
___________________
5 35 U.S.C. § 209(d)(3).
to PAEs do not appear to spur innovation (Feldman and Lemley, 2018). Although it is unclear how these exclusive licenses of federal technology served the public interest, promoted the dissemination and use of the patented inventions, or otherwise complied with the requirements of Section 209, courts have found that private parties cannot enforce the provisions of Section 209.6
There is no evidence that issuing licenses when the public interest is not served (such as licensing to PAEs) is a common occurrence, largely because no audit has been conducted to determine whether federal laboratories are issuing exclusive licenses that are in the public interest. Such an audit could be conducted by the Government Accountability Office (GAO) to determine the extent of the problem and continued on an ongoing basis to help provide information about future licensing. Such an audit would check to make sure that license agreements include strong downstream controls, such as licensing to others if the exclusive licensee is unable or unwilling to meet demand (Feldman et al., 2020).
While the public interest requirements regarding exclusive licenses apply to GOGO labs, Section 209 does not place similar requirements on government-owned, contractor-operated (GOCO) labs (although some agencies, such as DOE, have issued rules mimicking Section 209’s public interest requirements). Although it is assumed that licensing will produce welfare gains by, among other things, promoting the introduction of new products to the market and the diffusion of technical knowledge, an explicit public interest requirement would allow the GOCO labs to consider alternative avenues for commercialization that could include firms, researchers, entrepreneurs, and inventors not usually invited into partnerships or collaboration with the labs.
Perhaps more important, there appears to be no effective remedy for violation of the Section 209 requirements on the part of federal agencies. Such remedies as recognition of such a violation as constituting an affirmative defense to a claim of infringement by an exclusive licensee of a federal patent or as giving rise to a private cause of action for the violation would help ensure that exclusive licensing by federal labs was in the public interest.
___________________
6 For example, Reilly and Waxman (2016) describe an encryption technology patent that the Naval Research Laboratory (NRL) licensed exclusively to Network Signatures, Inc., a known PAE. During the 5 years following the granting of this license, Network Signatures initiated 89 patent infringement lawsuits against a range of businesses, including Wal-Mart, Harley Davidson, and Goldman Sachs. Most of the cases were settled for undisclosed amounts. When one defendant, Citibank, challenged Network Signatures’ exclusive license from NRL as noncompliant with Section 209, the court held that the Bayh-Dole Act does not provide for a private cause of action and dismissed the claim (Network Signatures, Inc. v. Citibank, N.A., 2008 W.L. 5216032, at *3 [C.D. Cal. Dec. 4, 2008]). Rather, the only avenue for enforcing the provisions of Section 209 appears to be the agency’s own administrative notice and comment process initiated prior to the granting of any such license (Southern Research Inst. v. Griffin Corp., 938 F.2d 1249, 1253 [11th Cir. 1991]). However, such an ex ante procedure requires vigilance and the expenditure of substantial resources by potential infringers of licensed patents.
Inventions Developed under a Cooperative Research and Development Agreement
Under the 1986 amendments to the Stevenson-Wydler Act, a federal laboratory that is a party to a CRADA may grant to its collaborating party an exclusive license to or assignment of any inventions made by employees of the lab. Federal labs that own inventions arising from a CRADA are subject to the restrictions concerning licensing of federal inventions that are enumerated in 35 U.S.C. § 209.
In contrast, no such restrictions constrain collaborators to whom rights of ownership over inventions have been granted ex ante, even if federal employees have contributed to those inventions. In principle, however, the lab retains the right to use, or have used, the invention for government purposes.7 Likewise, CRADAs must allow for the collaborating party to retain title to any invention made by its employees, subject to the granting of a nonexclusive license permitting the lab to use, or have used, the invention for government purposes.8 While no reporting requirements attach to patents on inventions owned by collaborating parties, the USPTO does require that the face of issued patents reflect the use of federal funding by contractors.
Other Federally Funded Inventions
Under the Bayh-Dole Act, entities that receive federal funding, including contractors of GOCO laboratories, are entitled to retain rights to inventions they make in the course of government-funded research.9 These entities are free to license such rights, with the provision that any exclusive licensee must agree that products embodying, or produced using, the licensed invention will be manufactured substantially in the United States.10 In contrast with the comparable U.S. manufacturing requirement applicable to federal inventions under 35 U.S.C. § 209, however, Section 204 authorizes the funding agency to waive this requirement if not commercially feasible.
Reporting on Patents by Federal Laboratories
Current reporting requirements, including patent reporting requirements, for federal laboratories are discussed in detail in Chapter 7. In that discussion, the committee reviews the significant empirical limitations associated with those
___________________
7 15 U.S.C. § 3710a(b)(1). Notwithstanding this statutory provision, the committee heard evidence that some contractors may nevertheless charge the agency for the use of this technology when an express acknowledgment of the government’s retained rights is not set forth in the relevant contract.
8 15 U.S.C. § 3710a(b)(2).
9 35 U.S.C. § 204(a).
10 35 U.S.C. § 204. In addition, some agencies have issued directives on exclusive licensing that follow the concepts in Section 209 that apply to GOGO labs. See, e.g., DOE Order 483.1B, “DOE Cooperative Research and Development Agreements,” December 12, 2016.
requirements and suggests mechanisms for addressing those limitations. This chapter focuses on one limitation on patent reporting that is specific to the current statutory structure, 15 U.S.C. § 3710(f), and associated regulatory guidance. Under the current NIST Guidance for Preparing Annual Technology Transfer Reports, federal labs are required to report annually on the patents they own and the number of CRADAs they have executed. However, there appears to be no mechanism for tracking patents, or any other innovation metric, associated with CRADAs. Given that both the academic literature and industry point to CRADAs as one of the most successful mechanisms for technology transfer and commercialization from the federal government (as discussed in Chapter 7 and other chapters), the absence of additional tracking with respect to CRADAs appears to be a deficiency of the current legal/guidance regime.
TRADE SECRETS
This section provides an overview of trade secret law and its intersection with the Freedom of Information Act (FOIA) for federal digital products.
Trade Secret Law
Trade secret law exists at both the state and federal levels in the United States. At the state level, almost all states have adopted some version of the Uniform Trade Secrets Act (UTSA) published by the Uniform Law Commission. At the federal level, the 2016 Defend Trade Secrets Act (DTSA), 18 U.S.C. § 1836 et seq., offers a similar civil cause of action.11 Under each of these statutory regimes, the owner of a trade secret can bring an action against a party that obtains, uses, or discloses a trade secret without the owner’s authorization.
As defined in Section 1.4 of the UTSA, a trade secret constitutes “information, including a formula, pattern, compilation, program, device, method, technique, or process, that: (i) derives independent economic value, actual or potential, from not being generally known to, and not being readily ascertainable by proper means by, other persons who can obtain economic value from its disclosure or use, and (ii) is the subject of efforts that are reasonable under the circumstances to maintain its secrecy.” Under the DTSA, information that can constitute a trade secret comprises “all forms and types of financial, business, scientific, technical, economic, or engineering information, including patterns, plans, compilations, program devices, formulas, designs, prototypes, methods, techniques, processes, procedures, programs, or codes, whether tangible or intangible, and whether or how stored, compiled, or memorialized physically, electronically, graphically, photographically, or in writing,” with the same secrecy and economic value requirements as those of the UTSA.12 Trade secrets can
___________________
11 In addition, the Economic Espionage Act of 1996, 18 U.S.C. § 1831 et seq., creates a federal criminal action for trade secret misappropriation.
12 18 U.S.C. § 1839(3).
include a range of confidential commercial and business information, such as customer and prospect lists, sales forecasts, marketing plans, manufacturing processes, ingredient lists, bills of material, component costs, proprietary algorithms, and the like. It is important to note, however, that trade secret law relates to commercial information and not to government classified information, which is administered and regulated under 32 C.F.R. Part 2001, “Classified National Security Information.”
Trade secret protection lasts as long as the holder of the information makes reasonable efforts to maintain its confidentiality. Thus, a voluntary public disclosure of information by its holder, whether in a press release or a product manual, generally eliminates trade secret status. An exception is made if the information is disclosed to third parties pursuant to an enforceable nondisclosure or confidentiality agreement. Trade secrets also can be used without liability by third parties that use proper means, such as reverse engineering or independent invention, to acquire them. Because the contents of a U.S. patent application generally become public 18 months after filing, inventions disclosed in patent applications generally lose trade secret protection. For this reason, trade secret law is often described as an alternative to patents as a mode of protection (Lemley, 2011).
Many firms utilize trade secrets to protect digital products, particularly when other forms of legal protection, such as patent and copyright, are unavailable. Thus, commercial data and databases, which are not protected by copyright in the United States, are often treated as trade secrets, as are proprietary algorithms, particularly because patent protection for pure algorithms is unavailable.
Computer software may also be treated as a trade secret, even when copyright protection exists as well. For example, a firm may consider customized software it developed and uses internally (e.g., for economic modeling or programmed trading) to be a trade secret. The trade secret status of computer software that has been publicly distributed has vexed courts and commentators for decades (see Dratler, 1985; Katyal, 2020). For purposes of this discussion, suffice it to say that the combination of restrictive software licensing agreements that require users to maintain the confidentiality of even mass market software and the difficulty of extracting intelligible source code from executable object code programs has resulted in general recognition of trade secret protection for the internal mechanics of most publicly distributed software programs. That protection extends to their source code, though not necessarily to their visible screen displays or output (1 Milgrim on Trade Secrets § 1.09[5][b]). Needless to say, open-source code software, in which source code is made freely available to the public, is not subject to trade secret protection.
Commentators have criticized the increasing use of trade secrecy to protect digital products. Trade secrets do not require any disclosure, and legal protection is available for as long as secrecy is maintained (Thomas, 2014). Trade
secrets may also stymie scientific disclosure and progress13 because in contrast with the patent and copyright systems, which are designed to induce dissemination to the public, a trade secret’s value lies in its secrecy, which forecloses public access to the knowledge. Additionally, using trade secrets may result in a loss of future innovative ability due to decreased labor mobility (Marx and Fleming, 2020) or decreased incentives for inventors (Contigiani and Hsu, 2019; Contigiani et al., 2018). In her econometric analysis, Williams (2013) concludes that the release of genomic data into the public domain by the HGP was associated with the generation of more commercial products relative to the release of genomic data kept as trade secrets (and licensed accordingly) by Celera, the HGP’s major private competitor (Williams, 2013). As Professor Katherine Strandburg explained to the committee,14 trade secret law is particularly problematic because it reduces the potential for others to build on existing innovations; it lacks certain moderating doctrines, such as fair use, that characterize copyright and, to a lesser extent, patent law; and it may compromise quality control, reproducibility, and accountability. According to Fromer (2019), moreover, technological advances suggest that one key moderating doctrine for trade secrets, the propriety of reverse engineering, may be less effective for digital products than for nondigital products because digital products are increasingly perceived as more difficult to reverse engineer as a result of developments in, among other things, cloud computing, machine learning, and automation.
Intersection of the Freedom of Information Act and Trade Secret Law for Federal Digital Products
FOIA15 was enacted in 1966 to promote open government by ensuring that members of the public have access to federal agency records. It requires each federal agency (excluding Congress, courts, certain military authorities, and others) to make its records available promptly to any person that requests them, subject to reasonable procedures and fees.16 Agency records, for purposes of FOIA, include both physical and electronic records.17
FOIA plays an important role in the dissemination and public use of federal digital products. Even though government works are free from copyright under Section 105(a) of the Copyright Act, there is no other assurance that federal laboratories will provide these works to the public. As one commentator notes, “the absence of copyright does not by itself make federal government information available for general use.…In theory, the Copyright Act and the FOIA work together to ensure public availability and unrestricted use of government data. The
___________________
13 Presentation to the committee by Brian Lally, DOE, December 5, 2019.
14 Presentation to the committee by Prof. Katherine J. Strandburg, March 3, 2020.
15 Pub. L. No. 89-487; 5 U.S.C. § 552.
16 5 U.S.C. § 552(a)(3)(A). But as noted by Reichman and Uhlir (2003), FOIA is itself an imperfect mechanism for accessing information, as the process is “time consuming and bureaucratic.”
17 5 U.S.C. § 552(f)(2).
two laws are complementary parts of policy that supports public access to federal information resources.”18
There are, however, numerous exceptions to FOIA disclosure requirements, including information that is classified, constitutes personnel or medical files, reflects agencies’ internal deliberative processes, or pertains to law enforcement.19 In addition, an agency is not required under FOIA to disclose privileged or confidential “trade secrets and commercial or financial information” that it has obtained from a nongovernmental third party.20 Thus, if a private party submits trade secret information to a federal agency pursuant to an application or investigation, the agency ordinarily will not be allowed to release that information in response to a FOIA request. While the trade secret exception provides some benefits (e.g., enabling the filing of patent applications before information is released to the public), it has been criticized for both its breadth and its ambiguity (Levine, 2011). In addition, if information is created by a federal lab working with a private collaborator under a CRADA, then under the Stevenson-Wydler Act, that information may be treated as a trade secret and excluded from release under FOIA for a period of up to 5 years after its creation.21
There is some uncertainty as to whether computer software developed by or for a federal agency (not operating under a CRADA) constitutes an agency “record” subject to FOIA, and little statutory guidance addresses this issue. In the only reported case addressing this issue that the committee could identify, a Federal District Court distinguished between “generic word processing or prefabricated software,” which does not constitute a “record” for purposes of FOIA, and software developed by an agency researcher in order to access and manipulate a particular database, which does constitute a “record” subject to FOIA (Cleary, Gottlieb, Steen & Hamilton v. Dept. Health & Human Services, 844 F. Supp. 770, 782 [D.D.C. 1993]). Today, the online DOD Freedom of Information Act Handbook (DOD, 2020) explains that “a record is the product(s) of data compilation, such as all books, papers, maps, and photographs, machine readable materials, inclusive of those in electronic form or format, or other documentary materials, regardless of physical form or characteristics,” which appears to be less directed to software. Thus, software may or may not be within the scope of federal works accessible to the public under FOIA, at least in the view of DOD.
Trade Secret Status for Federal Digital Products
As early as the 1990s, critics of the federal government’s open-access policies argued that FOIA has been an “impediment” to agencies’ technology commercialization activities, given that “a foreign competitor can come in
___________________
18Gellman, 1994, p. 1004.
19 5 U.S.C. § 552(b).
20 5 U.S.C. § 552(b)(4).
21 15 U.S.C. § 3710a(c)(7)(B). This treatment appears to have been proposed in GAO, 1988 at 3.
through a FOIA and ask a Federal laboratory individual to give information on their research in progress, which was never the intent of FOIA” (U.S. House, 1992, p. 53 [Statement of Assistant Secretary of Commerce for Technology Policy Deborah Wince-Smith]; see also Chandler [1991, p. 404]: “Another significant obstacle frustrating the Government’s ability to transfer computer software technology is the requirement of public disclosure pursuant to [FOIA].”) Thus, opposition to the release of government digital products under FOIA resonates with both the perceived need for exclusivity in federal technology transfer and the fear of foreign free riders.
In response, Section 801(b) of the 2013 National Defense Authorization Act, Pub. L. No. 113-66, was enacted to provide a special 5-year FOIA exemption for unclassified computer software and documentation developed at a DOD laboratory for military purposes.22 This FOIA exemption applies if the software would have qualified as a trade secret had it been obtained from a private party, effectively putting DOD-developed software on par with software developed by a private party under a CRADA (discussed above). As a result, this unclassified government software can be withheld from parties making FOIA access requests absent any traditional exemption from FOIA. Section 801(b) also authorizes the DOD lab to license the trade secrets in this software to third parties if, among other things, the public is notified of the availability of the software and related documentation for licensing, and interested parties have a fair opportunity to submit licensing applications, the licenses comply with the requirements of 35 U.S.C. § 209, and the software was originally developed to meet DOD’s military needs.23 Such licenses may be exclusive or nonexclusive. The first reported license issued under this authority was granted by the Air Force Research Laboratory to Pratt & Whitney and United Technologies Corporation in 2018 on a nonexclusive basis.24 Given the arguments raised above regarding the impact of trade secrets on innovation, it is clear that unless there is a national security issue, from a public interest perspective, trade secrets are a less desirable form of IP protection for federal labs’ digital products.
___________________
22 The provisions of Section 801 originally expired in 2017 but were extended through 2021 under Pub. L. No. 114-328, § 818. A representative of the Office of Naval Research told the committee that that agency hoped the provision would soon be made “permanent.” Presentation to the Committee, John Karasek, Supervisory Intellectual Property Counsel, Office of Naval Research, March 2, 2020.
23 As a condition for granting such licenses, the DOD lab must notify the public of the availability of such licenses and ensure that interested parties have a fair opportunity to submit applications for licensing. And if the DOD lab wishes to grant exclusive licenses to such software, it must comply with the public interest requirements under Section 209 of the Bayh-Dole Act.
24 “AFRL signs first of its kind software license with Pratt & Whitney,” Apr. 25, 2018, https://www.wpafb.af.mil/News/Article-Display/Article/1503043/afrl-signs-first-of-its-kind-software-license-with-pratt-whitney/.
INTELLECTUAL PROPERTY SURROGATES: CONTRACTUAL AND TECHNOLOGICAL MEASURES
As discussed earlier in this chapter, many valuable digital products created by federal laboratories lack formal IP protection. As discussed in the next chapter, with a few conspicuous exceptions, copyright does not exist in government works. In addition, patents are difficult to obtain for digital inventions, and FOIA eliminates the less desirable trade secret protection for unclassified agency information. Nevertheless, federal agencies seeking to incentivize commercialization and/or extract rent from the transfer of these digital products to the private sector can circumvent this lack of protection via two distinct routes. First, as described earlier in this chapter, they can continue to obtain a range of statutory, regulatory, and judicial exceptions to the baseline legal exclusions from IP protection for these works. Second, by imposing restrictive contractual terms and technological control mechanisms on these digital products, they can compensate for the lack of legal protection. These mechanisms act as surrogates for IP protection.25
Commentators and courts have long recognized the power of contract and access mechanisms to limit the use of digital products even without express IP protection through private contracts. Thus, it is relatively settled law that software license agreements may prohibit users from taking such actions as reverse engineering that would otherwise be permissible fair use under U.S. copyright law (Bowers v. Baystate Techs., Inc., 320 F.3d 1317 [Fed. Cir. 2003]).
Over the years, federal agencies have taken advantage of a number of mechanisms—including nondisclosure agreements, license agreements, restrictions on redisclosure of information, and other types of controls—to exert greater control over federal digital products.26 Such contractual limitations continue to be utilized today. Even without copyright, for example, the National Cancer Institute (NCI) distributes its NCIDOSE radiation dosimetry tools only in the form of executable object code and claims that “title in the Software shall remain with [NCI].” The software is released for noncommercial users under a free software transfer agreement that prohibits users from treating or diagnosing human subjects; using the software for commercial purposes; and copying, modifying, extending, or making derivatives of the software.27 Parties interested in licensing the tools for commercial purposes must enter into a license agreement
___________________
25 See Reichman and Uhlir, 2003, p. 381, “Electronic contracts become substitutes for intellectual property rights to the extent that they make it infeasible for third parties to obtain publicly disclosed but electronically fenced data without incurring contractual liability for damages.”
26 See Gellman (1994). The Digital Millennium Copyright Act (DMCA) of 1998, Pub. L. No. 105-304, 17 U.S.C. § 1201 et seq., prohibits the circumvention of any technological measure that protects a copyrighted digital work. The DMCA has often been cited as a significant legal barrier to the free use of public-domain data and other material (Reichman and Uhlir, 2003). Given the lack of copyright in government-created works, however, it is not clear that the DMCA is directly applicable to such works.
with NCI. These requirements appear designed to protect NCI from liability associated with use of the software.
FINDINGS AND RECOMMENDATIONS
Finding 4-1: Given documented cases in which federal laboratories have licensed software patents to patent assertion entities, it appears that federal labs do not always adhere to the requirements in Section 209 of the Bayh-Dole Act, which include allowing licensing only if such action is necessary to achieve practical application. However, evidence of how common such breaches are is limited. At the same time, there appears to be no effective remedy for violation of the Section 209 requirements on the part of federal agencies.
Finding 4-2: Exclusive software licenses are less valuable to the private sector if the software and source code are accessible via the Freedom of Information Act (FOIA). Using trade secrets to protect federal software, including source code, limits its accessibility to the public under FOIA.
Finding 4-3: Differences among federal agencies with respect to the trade secret exemption from the Freedom of Information Act appear to have no clear policy rationale.
Finding 4-4: Over the years, federal laboratories have taken advantage of a number of mechanisms—including nondisclosure agreements, license agreements, restrictions on redisclosure of information, and other types of controls—to extend their control over federal digital products, even in the absence of recognized intellectual property protection.
Recommendation 4-1: The Government Accountability Office or another appropriate, independent government entity should regularly, but not less than biennially, conduct a comprehensive assessment of the federal laboratories’ compliance with public interest requirements in Section 209 of the Bayh-Dole Act related to issuance of exclusive patent licenses.
Recommendation 4-2: Congress should consider enacting mechanisms that provide greater legal force to public interest licensing requirements. For example, the violation of such requirements could be recognized as an affirmative defense to a claim of infringement by an exclusive licensee of a federal patent or give rise to a private cause of action for such violation.
Recommendation 4-3: Congress should consider imposing public interest licensing requirements on government-owned, contractor--
operated laboratory contractors that are comparable to those imposed on their government-owned, government-operated counterparts under Section 209 of the Bayh-Dole Act.
Recommendation 4-4: Congress should consider rescinding the Section 801 authority provided by the 2008 National Defense Authorization Act that allows agencies to assert trade secrets even for unclassified information with no national security implication.
This page intentionally left blank.
