Statement of C. Paul Robinson, Director
Sandia National Laboratories
Mr. Chairman and distinguished members of the committee, thank you for the opportunity to testify today. I am Paul Robinson, director of Sandia National Laboratories. Sandia is managed and operated for the U.S. Department of Energy (DOE) by Sandia Corporation, a subsidiary of the Lockheed Martin Corporation.
Sandia National Laboratories is a multiprogram laboratory of the U.S. Department of Energy and one of three DOE laboratories with a research and development responsibility for nuclear weapons. Sandia's job is the design, development, and certification of nearly all of the non-nuclear subsystems of nuclear weapons. Our responsibilities include arming, fuzing, and firing systems; safety, security, and use-control systems; engineering support for production and dismantlement of nuclear weapons; and surveillance and support of weapons in stockpile. We perform substantial work in programs closely related to nuclear weapons, such as nuclear intelligence, nonproliferation, and treaty verification technologies. As a multiprogram national laboratory, Sandia also performs research and development for DOE's energy and science offices, as well as reimbursable work for other federal agencies (principally the Department of Defense) when our unique capabilities can make significant contributions.
I appreciate the committee's interest in the sale and buy-back of a surplus Intel Paragon computer by Sandia National Laboratories in 1998 and 1999. I shall endeavor in my statement today to faithfully summarize the circumstances surrounding those transactions and answer your direct questions on this matter. In addition, I will enumerate several issues that are raised by this incident and which, in my opinion, need to be dealt with.
Let me begin by dispelling two erroneous inferences about this transaction:
Sandia's excess Paragon computer was neither sold nor exported to the People's Republic of China; nor was it made operational by the buyer.
No classified information was released with or on the computer when it was transferred.
ACQUISITION AND USE OF THE PARAGON
BY SANDIA NATIONAL LABORATORIES
Sandia National Laboratories acquired the first Intel Paragon XPS supercomputer in 1993. The purchase price of approximately 10 million dollars was funded through a cooperative arrangement between the DOE Office of Defense Programs and the National Consortium for High-Performance Computing, sponsored by the Defense Advanced Research Projects Agency (DARPA). DARPA was an active sponsor of research into massively parallel processing, and in the early 1990s Sandia was one of the nation's leaders in the development of software tools for applications of this new computational technology. The National Consortium for High-Performance Computing included many organizations with a strong research interest in massively parallel computing, including several universities, DoD research laboratories, and the National Science Foundation's supercomputing centers.
The primary use of the Paragon XPS supercomputer by Sandia and members of the consortium was to advance the state-of-the-art in computational science. Sandia used the Paragon to develop and test new algorithms and software tools for massively parallel processing that would be useful in DOE applications. Researchers at the NSF supercomputing centers, various research universities, and the DoD research laboratories were also investigating parallel processing and were permitted to use the machine in its unclassified configuration.
The Paragon was occasionally used for classified processing related to DoD and DOE programs. Nuclear weapons applications run on the Paragon were principally calculations of weapon accident scenarios and simulations of shock wave impacts on internal components.
Ordinarily, Sandia's Paragon supercomputer was configured to be accessible to consortium members, such as NSF supercomputer centers and universities. Whenever the machine was to be used for classified processing it would be reconfigured in accordance with standard procedures specified in a security plan that had been approved by DOE. The machine, housed at all times in a limited-access classified vault, would be powered-down (a procedure which purges memory) and its communication lines disconnected from the unclassified network. Disk drives containing the operating system would be removed, and unclassified data storage would be powered-off. A second set of clearly marked classified disk drives containing a protected copy of the operating system would be inserted, and a second set of drives with space for classified data storage would be powered-on. The Paragon's communication lines would then be connected to a special classified gateway computer which controlled access from Sandia's classified network. Thus, access to the classified Paragon would be possible to only a few approved users residing on Sandia's classified network.
After these procedures had been monitored by a technician other than the one who performed the reconfiguration, the machine would be powered-up in its classified mode. Reconfiguration for unclassified processing would require a similar process. These security procedures guaranteed that all classified data was isolated to a separate set of storage hardware that could never inadvertently remain on the machine when it was in its unclassified configuration. These procedures also guaranteed that the system was not accessible to outsiders, nor to cleared employees without the proper need-to-know, when operating in its classified configuration.
SANDIA'S DECISION TO SELL THE PARAGON
In the spring of 1998, the Paragon XPS supercomputer was judged obsolete for Sandia's needs. Because of its experimental nature, the machine had a higher failure rate than was desirable (up-time ranged from 10 to 50 hours between interrupts) and was not sufficiently robust to be useful in a production environment. Moreover, the Paragon's capability had been surpassed by Sandia's Option Red computer under DOE's Accelerated Strategic Computing Initiative (ASCI).
Two-thirds of the Paragon system was disassembled, stored in Sandia's computer facility, and offered for sale in June 1998 under the federal exchange/sale provisions authorized in 41 CFR 101-46. Sandia retained the remaining one-third of the system, including all disks and drives that had been used for classified processing. The classified disk drives were subsequently degaussed and salvaged. The Paragon system was advertised to potential buyers on the open market. A single resulting bid of $2,550 was considered too low to proceed with a sale, and a decision was made to declare the system excess and offer it to other DOE sites and federal agencies.
No federal entity expressed interest in the decommissioned system. The computer was a six-year-old technology by 1998. It lacked the reliability required of a production system and cost about a million dollars a year to operate. Government agencies had access to more reliable computers with comparable computing power and lower cost of operation. Even Oak Ridge National Laboratory, which operated a similar Paragon, declined to take the machine for spare parts!
When no federal entity expressed interest in the decommissioned system, Sandia decided to re-advertise it for sale as a surplus item. It was the judgment of our computer experts that the machine might be useful as a research machine in a university computer science department or as a source of spare parts for another Paragon.
SALE AND RECOVERY OF THE DECOMMISSIONED PARAGON
On September 2, 1998, EHI Group USA, Inc., of Cupertino, California, contacted Sandia and expressed interest in the Paragon system. EHI claimed to own a Paragon (not necessarily the supercomputer version) but needed parts. Sandia formally re-advertised the property for sale to several potential bidders (including the bidder from the previous offering and EHI). EHI was notified of their high bid of $30, 800 on September 29, and the equipment was sold to them on October 26, 1998.
EHI took possession of the decommissioned Paragon computer components in Albuquerque on the following day. The shipper document, signed by Mr. Korber Jiang of EHI Group in the presence of the Sandia representative, was clearly labeled with the following standard export control notice:
ALL COMMODITIES MAY BE SUBJECT TO EXPORT CONTROLS UNDER THE CODE OF FEDERAL REGULATIONS, TITLES 10, 15, 22 & 31. ANY EXPORT OF THESE COMMODITIES CONTRARY TO U.S. LAW IS PROHIBITED. FOR ASSISTANCE CALL THE U.S. DEPARTMENT OF COMMERCE BUREAU OF EXPORT ADMINISTRATION. THE INFORMATION IN THIS PARAGRAPH REGARDING EXPORT REQUIREMENTS MUST ACCOMPANY ANY TRANSFER/SALE OF THESE COMMODITIES.
In November 1998, Mr. Jiang complained to Sandia that the computer system was not operational and requested additional parts. The Sandia representative reminded him that the system had been sold "as is, where is." He referred Mr. Jiang to the Intel Corporation for additional parts.
In early December 1998, a representative of Intel Corporation contacted a Sandia employee with concerns about EHI and Mr. Jiang. Mr. Jiang had contacted Intel to order additional parts, but since he could not provide details about what parts he needed, Intel sold no additional hardware to EHI. However, the conversations with Mr. Jiang raised suspicions with the Intel representative that Mr. Jiang may have exported the machine to a customer in the People's Republic of China-which was actually not the case. The Sandia employee who had been contacted by the Intel representative conveyed those suspicions to his manager and to the agent in charge of the sale. However, the agent judged that information to be mere speculation, and it was not brought to the attention of Sandia's senior management or export control and counterintelligence offices at that time.
It should be noted that in late 1998 there was no general alarm about the espionage activities of the People's Republic of China. The Cox Committee did not release its report, "U.S. National Security and Military/Commercial Concerns with The People's Republic of China," until January 3, 1999, and that report was not available in unclassified form until May 25, 1999.
However, general awareness of Chinese espionage activities became quite acute after the classified Cox Report was issued in January 1999. And in June 1999, the report of the President's Foreign Intelligence Advisory Board caused the DOE laboratories to reevaluate their security programs in light of the aggressiveness of the Chinese espionage threat.
In early July 1999, after we conducted DOE-mandated security immersion training at the laboratories (with particular emphasis on the espionage threat posed by the People's Republic of China), the Sandia employee again raised the issue of the status of the Paragon computer. This time, the matter was brought to the attention of export-control personnel. Sandia's Export Control office checked to see whether EHI Group USA, Inc., or its owner, Mr. Korber Jiang, were on the lists of companies or individuals with a history of export-control concerns. Those lists were:
Department of the Treasury, Office of Foreign Asset Control, List of Specially Designated Nationals and Blocked Persons
Department of Commerce, Bureau of Export Administration, Table of Denied Persons and Entities List
There were no indicators on either EHI Group USA or Mr. Jiang in these lists.
In mid-July, Sandia's representative telephoned Mr. Jiang to check on the status of the export-controlled material. Mr. Jiang said that he had not been able to make the computer operational and that it was still in his warehouse in California. During the conversation, the Sandia representative also learned that Mr. Jiang was a citizen of the People's Republic of China.
Sandia informed the local DOE office immediately. A team of Sandia and DOE personnel was formed within hours, and they consulted with the local office of the FBI and the Northern California office of the Department of Commerce Office of Export Enforcement. Personnel from DOE, Sandia, and the Department of Commerce inspected the property at EHI's storage facility in California on July 16. All the equipment appeared to be present and unopened, still sealed in the original plastic wrap used for shipping. Sandia was instructed by DOE to negotiate to buy back the equipment. Although EHI initially asked $1,400,000 for the property, we succeeded in negotiating a price of $88,888. Sandia regained possession of the equipment on July 21. The equipment is now stored at Sandia National Laboratories' site in Livermore, California.
Although it is relatively old computer technology, in theory the disassembled Paragon components could have been reassembled to provide a machine in the supercomputer class. If it had been successfully reassembled, made operational, and combined with appropriate computer software, such a computer could have been useful for nuclear weapon applications. More likely, it would be used to investigate the computer science issues associated with massively parallel processing. We believe, however, that the practical task of making the portion of the disassembled Paragon components operational would have been problematic. Moreover, other attractive options exist today for achieving supercomputer capability. Nevertheless, considering the uncertainties that had arisen about the true intentions of EHI Group USA, and our heightened security sensitivity with respect to the People's Republic of China following the Cox Committee report, the prudent course was to repurchase the material.
FINDINGS OF THE DOE INCIDENT REVIEW TEAM
The DOE Office of Contract and Resource Management and the DOE Office of Arms Control and Nonproliferation jointly performed a review of the Paragon incident and issued a report on September 23, 1999. That report identifies a number of findings concerning Sandia's actions. The committee undoubtedly has received a copy of the Incident Review Team Report.
The Incident Review Team found that Sandia had correctly followed the High Risk Property Procedures. However, the team found that we had not followed the DOE Guidelines on Export Control and Nonproliferation with the thoroughness that was desirable. The team also noted that Sandia had not obtain a waiver for transferring the Paragon's unclassified disk drives as excess property without first sanitizing them, and that Sandia had not performed an adequate inventory on the boxes of components (which had manuals inadvertently buried under computer cables). Although these variances are regrettable, they did not materially affect the sale, and the Incident Review Team found that the consequences of those oversights were not severe.
The DOE Incident Review Team made eleven specific recommendations that Sandia and DOE are addressing. We are observing a moratorium on reapplication of export controlled property pending revision of our policies and procedures and approval from DOE to resume that activity. Our new processes will require DOE agreement for the reapplication of any high-performance computing equipment. We are establishing an internal review process that requires the concurrence of Sandia's export control and classification teams before any items are placed in the reapplication pool for distribution beyond the federal government. In addition, we are incorporating new information on export control and nonproliferation requirements into the laboratory's security education program to increase awareness of these important issues throughout the staff.
ISSUES RAISED BY THE PARAGON INCIDENT
There are certainly lessons to be learned from this episode that apply to the broader question of effective export control. The Paragon incident discloses a conflict between the current legal and contractual requirements governing laboratory disposal of property and the broader issue of national security and counterintelligence. Sandia is required by its prime contract (Article I-84, incorporating DEAR 970.5204-21) to follow the directives of DOE in the disposition of excess property. This includes requirements to follow federal and DOE property management regulations governing such dispositions. The regulations clearly are intended to maximize the revenue realized from excess property but are vague about the responsibility and discretion of contractors for withholding sensitive property from sale.
Our experience in this sale, therefore, suggests a need for additional or modified guidance. For example, a new regulation requiring destruction of government property that is export controlled, instead of offering it for sale, might close the gap that contributed to the present situation. In this regard, reconciliation of the Commerce and State Department regulations (EAR and ITAR) with the DOE High Risk Property Regulations would be extremely helpful.
Another issue is to what extent the citizenship or resident status of a buyer or agent may legitimately be considered in sales of export controlled property. In the absence of new regulations to clarify this issue, the nation's civil rights laws (particularly section 1981 of the Civil Rights Act of 1870 as amended in 1991) probably limit our discretion to refuse to sell to U.S.-incorporated companies because of the citizenship or resident status of their officers, owners, or employees. One alternative might be to enact a new regulation directing DOE contractors (and perhaps other sellers) to exclude the sale of export controlled material to U.S. companies that may be detrimentally influenced by a foreign power or citizen of a foreign country. A regulation that specifically prohibits export controlled information to U.S. companies under the direction or strong influence of citizens of sensitive countries (the U.S. State Department designator) should be considered. A regulation crafted similar to the existing Foreign Ownership, Control, or Influence (FOCI) requirements (DEAR 952.204-74), may allow us to exercise appropriate discretion without liability.
It is very worrisome to me that we have limited discretion in how we can treat permanent resident aliens (PRAs) in export control transactions. If Mr. Jiang had been granted permanent resident status, we would have had no basis for notifying the Department of Commerce of our concern as we did in July 1999. Export regulations constrain us to treat PRAs the same as U.S citizens. In contrast, the intelligence community may exercise discretion to regard permanent resident aliens as non-U.S. citizens in counterintelligence matters. It may be useful to revisit and amend the regulatory guidance governing our obligations toward permanent resident aliens in export control transactions.
In retrospect, we regret the sale of portions of a decommissioned Paragon supercomputer in 1998 to a company whose owner was a citizen of the People's Republic of China. I hasten to point out, however, that (a) the initial press stories bear little resemblance to the facts, and (b) no harm to U.S. security resulted from this incident. As you know, the portion of the computer that was sold was never made operational by the buyer, nor was it exported.
I assure you that since the issuance of the Cox Report and the performance of security immersion training for 100 percent of our staff, we have become much more attuned to the Chinese effort to acquire export controlled, high-performance computer technology. Today I believe we would not have made such a sale, but would have attempted to seek an exemption on its disposal. It is unfortunate that this single incident leaves an impression of Sandia National Laboratories that I know is quite uncharacteristic of the culture and performance of our people and organization.
Excluding the exceptions noted by the Incident Review Team, Sandia complied with the relevant property management and export regulations for selling high-risk and export-controlled property to a U.S. company. Consequently, a reexamination of existing policy and regulations governing contractor responsibilities and the discretion which could be exercised in such transactions would be useful. We are eager to cooperate with DOE, the Department of Commerce, and other responsible agencies to identify and resolve conflicts between property disposition requirements and national security interests.