On January 2, 2002, President Bush made it easier for the U.S. computer industry to sell high performance computers (HPCs) to China, Russia, Israel, India and Pakistan and 52 other so-called Tier 3 countries. These are countries of concern to U.S. national security interests. China has an advanced nuclear weapons program and it has helped Pakistan develop nuclear weapons and missiles. Relaxing controls on the export of our HPCs is likely to assist China in expanding its own nuclear weapons capabilities and their proliferation. The White House news release said the President’s decision to ease restrictions on the export of HPCs would protect “truly sensitive goods and technologies” and “promote national security.”
This decision marked the seventh time since 1993 that export control thresholds on HPCs to Tier 3 countries had been raised. Six of these occurred during the Clinton administration. Overall, they have increased by nearly a thousand fold the speed and power of computers that may be exported to Tier 3 countries. Clinton’s decisions sparked furious debate on Capitol Hill over the potential harm to U.S. national security from such sales, particularly to China and Russia. Many characterized his actions as a payoff for campaign contributions from the computer industry to the Democratic National Committee.
Despite allegations that President Bush’s decision was in response to a plea from an industry-lobbying group and the fact that the Bush campaign received twice as much in PAC and soft money from the computer industries as did the Gore campaign, there was virtually no reaction to it. Both the media and Capitol Hill were relatively silent.
By 1998, Congress was sufficiently concerned to impose, through the National Defense Authorization Act (NDAA), new reporting requirements on the administration each time it sought to raise export control thresholds on computers. Clinton did so one last time, on January 10, 2001, ten days before he left office. The announcement was accompanied by a proposal to eliminate controls on computer exports altogether. President Bush’s 2002 decision raised the export control threshold over two times higher than Clinton’s last act. In contrast to Clinton’s actions, however, Bush’s announcement largely escaped criticism until August 2002, when the General Accounting Office criticized it in terms strikingly similar to those used in earlier GAO evaluations of the Clinton administration’s actions on HPCs.
The General Accounting Office issued a report in which it graded the 2002 Bush announcement against three key 1998 National Defense Authorization Act reporting requirements. The GAO found the report to be lacking critical details and that, overall, it failed to offer an adequate justification for the decision. First, the GAO charged that the President’s report to Congress failed to include an assessment of the foreign availability of HPCs capable of operating at comparable performance levels. Instead, the administration relied on data provided by the Computer Coalition for Responsible Exports, a Washington-based lobbying group representing all the big names in the computer industry. The Computer Coalition told the administration that computers performing at the new threshold would be “widely available through foreign and domestic companies by early 2002,” implying that the U.S. could be shut out of this market.
But the GAO independently surveyed ten companies, cited in the administration’s report. It found only one producing the new computer system-Unisys Corporation, which co-chaired the Coalition. Unisys had channeled all of its 2000 campaign contributions to the Republican National Committee. When the GAO queried potential foreign producers, like NEC, Bull and Hitachi, these companies reported no plans to either produce or market HPCs in this class.
The administration was supposed to provide an assessment of potential uses of these HPCs that would be of “military significance.” But it simply stated that such computers are used in virtually all military and national security applications. Defense officials told GAO that such a list would be of “questionable” value and that the President’s decision was “driven by the market and what the administration believes it can control, not by the military and national security applications that could be run on high performance computers.”
Finally, the GAO determined that the President’s report provided no assessment of the impact on U.S. national security of China or Russia or any other Tier 3 country obtaining such computers, as required. Instead, the report claimed that “high performance computers would be of little or no value to countries of concern not having the requisite knowledge and experience in using these computers to advance their military capabilities.” Clearly, this does not include China or Russia, as the government has acknowledged some time ago.
In March 2001, the GAO issued a similar report criticizing the Clinton administration for its January 2001 relaxation of export controls on HPCs. The GAO criticized that report for its feeble attempts to address all three reporting requirements and especially for the omission of any reference to the potential military uses by China. That GAO report was overshadowed by other controversies surrounding Clinton’s last days in office. China seems to have been a problem for both the Clinton and Bush administrations.
The debate over export controls on high performance computers has come to symbolize the tension between globalization and the requirements of U.S. national security. During the Cold War, the U.S. and its allies were able to agree on restrictions on exports of critical military and “sensitive enabling” technologies like HPCs. But with the advent of the information revolution, more and more U.S. defense technologies came from non-defense industries, whose profits depend on sales to the civilian sector of the economy and, increasingly, penetration of world markets.
For many, and especially the computer industry, China now seems to represent the mother of all foreign markets. Industry representatives say that China is the world’s largest market for cell phones and the third largest for personal computers and claim that the future success of information technologies companies is tied to the successful penetration of Chinese markets. The common refrain is that if the U.S. won’t sell to the Chinese, others will-like the Japanese or the French. Industry representatives, and their supporters in government, also argue that U.S. national security will be harmed if these firms are restricted in their ability to compete in China or Russia. The industry is heavily dependent on investment in research and development. Restrictions on marketing to China will hurt the profitability of these companies and, consequently, drive down their R&D investments. These investments keep the industry on the cutting edge of technology, which benefits directly the military, the intelligence community and other national security functions.
Finally, the industry argues, and the GAO agrees, that the current metrics used by the government to measure performance are irrelevant and out-dated. Industry lobbyists prefer no restrictions on computer exports. They are supported in this by assessments done by the Energy Department’s nuclear weapons labs. Lab studies, done over the past several years, claim that export controls provide a “false sense of security” and should be eliminated.
Critics of these claims charge that industry and their supporters in government, play fast and loose with data used to support their case. This is reinforced by the GAO’s recent finding that, contrary to industry’s claim, only one company based in the U.S. is producing systems at performance levels permitted under the Bush decision. In fact, the history of this debate shows that proponents of loosened controls have consistently relied on studies, especially those performed at Stan-ford University, that are flawed and incomplete. These studies have consistently overstated the availability of foreign com-petitors and assumed, incorrectly as it turns out, that they would rush into new markets like China. However, the main competitor, Japan, has demonstrated far greater discipline in its export controls than anticipated. In 2001, for example, Japan, which now makes the world’s fastest supercomputer, rang up no sales of the faster HPCs or supercomputers to any Tier 3 countries, including China.
One of the most commonly cited Stanford studies, done in 1998, purported to examine national security applications on HPCs and the value of export controls in the future. But the study’s authors admitted later that they lacked sufficient information to assess which countries could use HPCs for military applications. Bush administration officials told a U.S.-China security review panel in mid-January 2002 that they were awaiting the outcome of a U.S. intelligence community study before making any further decisions on export controls. That study had already been done in 1998 by the Energy Department and the Defense Intelligence Agency. It was evident at that time that the CIA was poorly equipped to make such a study and that it may not even have collected the data needed to make informed judgments.
The United States maintains its preeminence as a military power based in part on the strength of its defense science and technology. It has used ever-higher levels of computing power to stay ahead of the rest of the world in terms of next-generation weapons and intelligence systems. Earlier than most, U.S. defense planners realized that advantage in future warfare will be based on the ability to conduct battlefield operations that are faster and more lethal than an opponent. U.S. defense industries use HPCs to develop weapons systems that fire at targets over the horizon with great precision and high lethality, as demonstrated in the Gulf War and again in Afghanistan. High-speed computers provide battlefield commanders better information more rapidly, which enables them to react faster to emerging situations and threats. Reliance on high performance computing has saved the U.S. billions of dollars in R&D costs and, equally significant, time in comparison to its potential adversaries.
Not surprisingly, these potential adversaries-like China-have sought to acquire similar capabilities. China tried to develop such capabilities, creating a knock-off of an IBM System 360, one of the earliest U.S. general-purpose mainframe computers. The Chinese system failed to meet performance expectations, however, and in the early 1990s, the Chinese began dangling the prospect of a vast new market in front of U.S. computer companies. By 1995, the Chinese were able to begin importing HPCs of increasing power, thanks to the Clinton administration’s relaxation of export controls. In little over a year, from January 1996 to March 1997, China imported forty-seven HPCs from the U.S. By 1999, this number had grown to 603. In 2001, U.S. companies shipped 112 HPCs to China, including six operating at or near the currently permitted performance level. The six to China represented nearly half of all such licenses issued in 2001. Each time the Clinton administration raised control thresholds, it would cite the rapid advances in computing technologies and the market pressures on U.S. firms as the rationale for its decision. The value of these exports was estimated to be no more than 500 million dollars, less than one per cent of U.S. exports to China.
In 1998, the administration negotiated an agreement with the Chinese to provide for U.S. inspections of end users to ensure that HPCs were not being diverted to military purposes. This was important, in part, because the administration had shifted the burden onto the U.S. supplier to ensure that the end user was legitimately civilian and that the exported systems would not be transferred to military users. End user inspections did produce one early success when a computer was found to have been diverted by China and was eventually returned to the U.S.
Irrespective of the appeal of its supposed “market,” there are tangible national security risks associated with permitting sales of HPCs to China. While there has been sharp debate over China’s future relations with the U.S. since the mid-1990s, no one doubts that China is emerging as a global economic power and a strong regional military power. While the China lobby in Washington and elsewhere promotes closer ties, expanded commercial relationships and greater Chinese access to U.S. science and technology as a way to improve overall relations, others doubt that U.S. and Chinese national priorities and interests will converge over time. Much of this debate crystallized over proposals to export HPCs to China.
This is because HPCs are useful for a range of military applications of interest to the Chinese, like in the development of modern air defense systems, advanced cruise missiles or improvements in ballistic missiles. The U.S. Defense Department has identified over 300 specific military applications that require HPC power to run. But it is in the area of nuclear warhead design and development that overall computing power is critical.
In 1998, the Energy Department’s intelligence office was tasked by the Congress to develop an assessment of the risks to U.S. security resulting from previous exports of U.S. HPCs. Energy was to look at the implications, world-wide, of these exports for foreign nuclear programs. After a long tussle with the Commerce Department, which refused to release data on U.S. exports necessary for such a study, Energy determined that of the countries surveyed, China would benefit most from the acquisition of U.S. HPCs. The more computing power available to the Chinese the greater the potential benefits.
This conclusion was based on a number of factors, including China’s relatively modest nuclear weapons expertise in comparison with the other major members of the nuclear club, the U.S., UK, France, and Russia. The Chinese program had lagged the U.S. and Russia by decades, but the Chinese had made great strides beginning in the early 1990s, assisted in part by a campaign of both overt and covert (espionage) acquisition of nuclear knowledge from the U.S. nuclear weapon labs. The Chinese had leaped ahead in the sophistication of their nuclear warhead designs and had finished an underground series of tests on new classes of nuclear warheads by 1996. At that point, the Chinese government announced its intention to sign the Comprehensive Test Ban Treaty and imposed a moratorium on all future testing of nuclear weapons.
Despite our belief that the Chinese had successfully tested new nuclear warheads, there were still many problems confronting the Chinese with regard to their lack of experience with these warhead designs. In particular, the Chinese had no experience with the aging process of the nuclear materials contained in these new warheads and the potential effects on performance, reliability and safety. If they could no longer test underground, Chinese scientists would have difficulty completing their modernization program. But there was an alternative-one that required them to shift the focus of their espionage campaign against the U.S. nuclear weapons labs.
U.S. scientists had come up with an alternative to underground testing to monitor the reliability of nuclear warheads, some of which by this point were reaching the end of their life expectancy. In the past, the labs could take a live weapon from the stockpile and shoot it off at the Nevada test range. These underground tests were extensively instrumented and scientists spent months analyzing data seeking to uncover flaws or imperfections that might have resulted from the aging of nuclear materials in the warhead. But the Clinton administration’s outright ban on live testing forced the labs to rely on computer simulations to replicate the results achieved from underground shots. To do this, the labs would have to develop more powerful computers and update the computer codes and simulation tools that had been used to design new nuclear warheads.
On the computer front, the labs took different paths. Each of the three nuclear warhead design labs teamed with a computer industry giant, like IBM or Silicon Graphics, to develop more powerful supercomputers. These partnerships produced four of the top twenty fastest supercomputers in the world. Simultaneously, the labs undertook to update their nuclear weapons codes and simulation tools to run on these computers. This was known as the “Legacy Codes” project. The objective of the overall program was to develop sufficient computing power that would enable the labs to run their updated codes, which replicated the performance of a warhead when detonated, as fast and as comprehensively as possible. Reducing the run time of the computer simulation programs would increase the productivity and efficiency of the overall program and, in the long run, save money and time.
Another study done by the Energy Department in 2000 determined that HPCs operating at speeds on the order of 10,000 million theoretical operations per second (MTOPS) or better “would be of significant use to China’s designers in examining likely gaps in their nuclear weapons programs.” U.S. export control thresholds passed that milestone early in 2000; by August 20002, the thresholds were up to 28,000 MTOPS. Obviously, faster is better and raising the threshold ever higher simply eases the burden and cost imposed on the Chinese nuclear warhead development program by enabling them to procure HPCs previously unavailable to them.
Ironically, one of the proponents of eliminating controls altogether made the case for protecting not only HPCs but also computer codes and simulation tools. James Lewis is the Director of Technology Policy at the Center for Strategic and International Studies and a former State Department arms control negotiator. Lewis ran a major study supporting government efforts to modernize export controls. During testimony before the U.S.-China Security Review Commission in January 2002, he dismissed the importance of computing power and said that without “sophisticated computer codes based on extensive experience and test data” HPCs would be of little value to weapons designers.
Lewis testified: “For nuclear weapons design, a central concern in the computer export control debate, access to data derived from nuclear weapons explosions is more important than computing power. A country without extensive experience in weapons design is at a significant disadvantage, and the lack of reliable data and proven codes will substantially constrain the usefulness of computing technology for military or proliferation purposes.” Lewis was obviously aware of China’s limited nuclear weapons history and was asserting that lack of experience would deny its scientists the ability to exploit HPCs to improve their nuclear arsenal, no matter how powerful those computers were.
The simple fact is, however, we don’t know whether the Chinese have acquired “reliable data and proven codes.” In the nearly forgotten Congressional Cox Committee report, the committee warned that the U.S. Legacy Codes contained precisely the type of data sought by the Chinese and there were “no procedures in place that would prevent or detect the movement of classified information, including classified nuclear weapons design information or computer codes, to unclassified sections of the computer systems at U.S. national weapons laboratories.” The committee feared that the Chinese, other government-sponsored computer hackers, or even terrorists could penetrate these highly vulnerable “unclassified sections” and obtain the codes.
What the Cox Committee did not know at the time was that these codes and their associated files had indeed been placed on the unclassified section of the Los Alamos computer network. Be- ginning in 1988, Wen Ho Lee had systematically created and then transferred computer files representing a “portable nuclear weapons design capability” onto the unclassified Los Alamos network. Forensic computer assessments of his actions in 1999 determined that he had stored nearly 800 megabytes of classified data, about the equivalent of one and a half compact discs, on the unclassified network and portable tapes. Los Alamos computer logs recorded over 300 unauthorized entries into the network in the space of about twelve months alone in 1997-1998. Lab scientists told the government in late 1999 that these data could be installed and run on a supercomputer or computers of even less capability, like the HPCs shipped to China throughout the late 1990s.
A previously unpublished CIA damage assessment concluded that if China had obtained these data, “China would likely continue pursuing its existing designs, which have been enhanced by information already obtained from the United States, although the information on these tapes would provide China more detailed design information and more robust tools for designing weapons, which could help with a future, smaller advanced weapon.”
Although the Chinese agreed to permit on-site inspections in 1998, these safeguards simply have not worked. Michael Garcia, an Assistant Secretary of Commerce who runs the Export Control Enforcement program, testified in January 2002 that in the previous twelve months, his bureau had conducted 42 such checks. However, most of these were on systems at low levels of performance. He cited the example of the Chinese permitting the check of a computer system at a travel agency, but he testified that the Chinese had denied “strategic checks,” that is, inspections of higher-end systems. As of last January, there were more than 700 U.S. HPCs in China awaiting inspection. We don’t know if or how many of these are now performing militarily-significant applications.
In August 2000, the government offered Wen Ho Lee a plea agreement. The government had indicted him on 59 counts of mishandling classified information, but now would allow him to plead guilty to just one count, and his sentence would be time already served and a small fine. In return, he would reveal what he had done with the computer tapes. He accepted the offer and the government has honored it despite the fact that the government still doesn’t know what he did with the tapes. He told the FBI that he had thrown them into a dumpster at Los Alamos National Lab. FBI agents searched the dumps, but, to no one’s surprise, they failed to find these tapes that bore classified information of great value to China.