Thank you, Mr. Chairman, for the opportunity to discuss once again export controls on high performance computers. I would like to discuss the evolution of control levels, our safeguards program, HPCs in China and Russia, and viable control levels.
The Bush Administration controlled computer exports to most destinations at the level of then-commonly-available personal computers. A supercomputer was defined as performing at more than 195 MTOPS. Today, many if not most PCs exceed that level. When President Clinton took office, Congressional leaders of both parties urged reform of this clearly outdated standard. An initial reform in February 1994 raised the control level to most destinations to 500 MTOPS and the definition of supercomputer to 1500 MTOPS, but these thresholds were soon overtaken by advances in computer technology.
In January 1996 the Administration again adjusted controls to reflect availability and national security relevance. That decision took into account work done by the Stanford University Center for International Security and Arms Control and several agencies, including the Department of Defense. This is the structure the Committee is familiar with -- MTOPS thresholds of 2,000, 7,000, and 10,000 depending on the country tier. For Tier 3 countries, military end-users were distinguished from civilian end-users by a lower control threshold.
Most of the discussion since then has concerned Tier 3 countries, to which computer systems between 2000 and 7000 MTOPS for civilian end uses may be exported under license exception. Individual licenses are required above 7000 MTOPS for civilian use and above 2000 MTOPS for military and proliferation related end-uses/users.
Numerous safeguards underpin this policy. A license is required for the export of a computer of any level, regardless of destination, if the exporter knows it will be used in a proliferation-related activity. This is required by the Enhanced Proliferation Control Initiative (EPCI) and our regulations governing nuclear end-uses and end-users. Last year we established a process for publishing lists of end-users involved in the development of weapons of mass destruction. We have been publishing the names of such entities since February and expect to continue doing so. These lists put exporters on notice that exports to these users require a license.
Finally, a record-keeping requirement specific to exports of high performance computers compels maintenance of export transaction details, and a comprehensive outreach program. Through that program, BXA personnel meet with industry representatives, alert industry to special problems, seek their cooperation, help them know their customer, educate them on the legal requirements, and share information.
Use of these safeguards has been rigorous. We maintain a watch list against which every export license application is screened to clear all end-uses and users, and we conduct pre-license checks and post-shipment verifications when appropriate.
Following reports of certain Tier-3 countries' acquisition of HPCs for end users that would, in our judgment, require an individual license, we have further stepped up enforcement efforts. As part of this process, we are analyzing transaction details from all U.S. companies identified as producers/exporters of HPCs.
The producers reported exporting 1437 systems totaling $785 million between January 1996 and March 1997. Of those, 91 were exported to Tier 3 countries. Of those 91, 47 went to China and 10 went to Russia. These are the statistics I have shared with the Committee previously.
BXA has opened four investigations on HPC exports: two involving China and two involving Russia. All these investigations are in the hands of the Department of Justice. China has cooperated with us, and the Sun Microsystems computer was returned to the United States on November 9. The Russian officials have indicated that Russia is considering moving the computers in question to facilities where unclassified research is being done. We hope to continue discussions regarding the details in the near future.
Thus far in 1997, BXA has requested 22 post shipment verifications of HPCs exported to Tier 3 countries. Seventeen have been completed, all with favorable results. The remainder of the requests are pending at our embassies.
Our previous work has suggested that it is impossible to control the spread of computers below the level of 4000 MTOPS, and that, in any case, below 7000 MTOPS the speed of the computer makes little difference in weapons design. Iraq did not need HPCs to build its military arsenal, and computers used to make some of our most sophisticated weapons are far below the 2000 MTOP level. To catch computers used in weapons design, we would have to set levels that control the PCs you can buy at your corner store. But since they are also made by the thousands in Taiwan, Korea and China, even that unilateral step would be ineffective.
In 1995 we determined that at that time the lower boundary of controllability was between 4000 and 5000 MTOPS, rising to 7000 this year. The Wassenaar Arrangement members have agreed to decontrol all computers up to 2000 MTOPS and to require reporting only for computers above 4000 MTOPS.
Those levels become even more questionable when we look at technological developments. You are all familiar with "Moore's Law," which says, in effect, that the capabilities of the microprocessor chip double every eighteen months. Chips the size of a dime are now capable of operating at 1300 MTOPS, and industry is building and will begin to market next year a single chip capable of 2000 MTOPS. At least one company plans to offer laptop computers capable of 2000 MTOPS within the next year. I would not be surprised within a year or two to find that the 2000 MTOP level catches high performance video games, which need quick processing times for their advanced graphics.
Thus, as I have said before, our conclusion is that computer controls must be related to reality. However much we would prefer it to be otherwise, we must not delude ourselves: we cannot control the uncontrollable. The widespread availability of microprocessor and computer capabilities, compounded by computing power aggregation technologies such as networking, clustering and parallel processing, make attempts to control at increasingly high levels fruitless. Trying to do so undermines the credibility of our export control regime and diverts scarce enforcement resources that could be better spent on matters more directly related to our national security.
Traditional "specially designed" defense systems were easier to protect than today's emerging dual-use civilian-military technologies because the owners of the technology were fewer and the markets more limited. The conversion of military technology to civilian use and its consequent diffusion into world markets has made security controls more difficult. The more recent trend toward conversion of civilian-commercial technology to military use has compounded the problem.
As Mr. Bryen has stated in his paper, "Technology Security and the Revolution in Military Affairs," cross-over dual-use technology derived from commercial off-the-shelf technology requires widespread availability to be commercially viable -- and the military needs that viability if the technology is to be available to it as well. The technologies involved demand large capital investment and expanded markets to provide the necessary return on that investment to support further research and development on next generation products. Export controls work counter to these needs. They limit market expansion and consequent return on investment at the very time our own national security is increasingly dependent on the strength of these domestic technologies. Controls that do not keep pace with the rapid change in technology harm our long-term national security objectives.
Finally, Mr. Chairman, I would be remiss if I did not take this opportunity to restate the Administration's opposition to the computer language contained in the defense authorization. It sets forth a control regime which is unrealistic, both philosophically and procedurally.
Let me be a bit more specific about the Congressional language.
In April of 2002 the Bureau of Export Administration (BXA) changed its name to the Bureau of Industry and Security(BIS). For historical purposes we have not changed the references to BXA in the legacy documents found in the Archived Press and Public Information.