TESTIMONY OF

DR. STEPHEN D. BRYEN

Mr Chairman, I welcome the opportunity to testify before the Military Procurement Subcommittee on the subject of the sale or transfer of supercomputers to foreign entities or governments.

The sale or transfer of supercomputers is, and has long been, a sensitive national security issue. It is question that not only directly affects the United States, but is of great importance to America's friends and allies. Ultimately, it is a subject that affects international security and world peace.

My expertise is in technology policy. Technology policy considers how to enhance technology and America's technology leadership, and also how to prevent the loss of technology that is highly leverageable and could cause us strategic or economic harm. In my years of service in the Defense Department as the Deputy Under Secretary of Defense for Trade Security Policy and as the founder and first director of the Defense Technology Security Administration, I was closely involved in the issue of safeguarding supercomputers. I helped negotiate and implement the 1986 U.S. Japan Supercomputer Agreement, which set up a system whereby the two major producers of supercomputers agreed to carefully monitor and regulate their sale to third countries. This cooperation yielded significant security benefits and demonstrated that two highly competitive countries could work out an effective means to regulate trade in this sensitive equipment.

The U.S. Japan Supercomputer Agreement was put in place primarily to guard against nuclear proliferation. At the time, while we still had the Paris-based Coordinating Committee on Export Controls, known as COCOM, there was no control system for non-Communist countries. Both American and Japanese policy makers recognized that the transfer or sale of supercomputers could not be left to national discretion; the issue was of such sensitivity that coordination was needed between countries manufacturing supercomputers, before the technology would be released to a third country.

In this connection, I understand that when the U.S. decided to massively liberalize its controls on supercomputers starting in 1993, there was no prior consultation with Japan nor did we seek agreement from the Japanese government on this matter, although prior agreement is clearly contemplated in the U.S.-Japan Supercomputer Agreement.

One of the issues the Subcommittee will need to look into is whether supercomputers remain a strategically sensitive commodity. For the most part, the administration has decided that only a very limited subset of supercomputers would qualify as strategic. And even those are under an incredibly weak control system that cannot effectively safeguard the transfer of these machines to third countries.

The argument made by some is that supercomputers are not so important since nuclear weapons and their delivery systems can be designed with lower speed computers. They point out that many of America's nuclear weapons and delivery systems such as ballistic missiles and long-range bombers, were built on computers whose performance is inferior to the supercomputers of today. But this is wrong. America needs supercomputers to design the next generation of defense systems, reduce costs and improve performance, and make sure we stay number one. And it is wrong because supercomputers make it possible to do effective design engineering with less risk taking, less expensive and dangerous testing, and to increase the safety of nuclear weapons and other systems including ballistic missiles and smart weapons.

This is an issue, I might point out, that was well understood by the Department of Defense and the Department of Energy, and supported by both the State Department and the Commerce Department, in the 1980's. The opinion was then that supercomputers were a critical commodity and high on the list of the most sensitive national security technologies.

In January, 1986 the Department of Energy Office of Military Application commissioned a study by its three top national laboratories at Los Alamos, Lawrence Livermore and Sandia titled "The Need for Supercomputers in Nuclear Weapons Design." The study found that "The use of high-speed computers and mathematical models to simulate complex physical process has been and continues to be the cornerstone of the nuclear weapons design program [of the United States]." This study still stands as the definitive word on the use of supercomputers in designing nuclear weapons systems.

The Energy Department study considered the issue of "efficiency." Thanks to supercomputers, a new nuclear weapons design or concept involves exponentially fewer explosive tests. For example, in 1955 a new concept would require 180 tests; in 1986 the number of tests required was reduced to 5. As even more powerful supercomputers are available today, it is highly probably that the number of tests may be reduced even further, or testing altogether eliminated.

What does this mean? It means that a country that gets supercomputers can develop nuclear weapons covertly, and have plausible deniability if challenged. It means that we may totally misjudge the capabilities of a hostile country or potential adversary, as we did in the case of Iraq.

It also means that the cost of developing nuclear weapons is significantly reduced if supercomputers are available. This is an important point, because many countries lack both the requisite number of technical experts and the infrastructure to develop nuclear weapons. If it were otherwise, does anyone doubt that many more countries would be fielding nuclear weapons today?

We also need to ask, in respect to supercomputers that might go to countries that already have nuclear weapons, such as Russia or China, whether they can gain any benefit from getting them. For Russia and China the acquisition of supercomputers is of great importance in allowing them to develop a viable nuclear strike capability.

The Energy Department study makes the following important observation:

"Computer analysis is our link between the underground test and the wartime situation. A vehicle containing a nuclear warhead might be expected to withstand very high accelerating forces and warhead surface temperatures, which are changing rapidly as the warhead reenters the atmosphere. In addition, upon detonation, warhead output will be affected by background environment which is radically different underground (e.g., neutron reflection). These conditions obviously do not exist, and cannot be created, underground. They must, therefore, be simulated by computer. Computers enable us to infer real-environment weapon performance from underground tests.

The battle environment itself can also have an important influence on the performance of this warhead. In a multiple-warhead attack, each reentry vehicle is to some extent exposed to the blast and the radiation effects of many of the other warheads in the salvo. To protect against 'fratricide,' the designed must understand its effect on system performance. Since the true environment can only be produced in an actual attack, and practical constraints limit the extent to which the appropriate conditions can be produced underground or in above-ground experimental simulation facilities, the computer becomes absolutely essential in the evolution of a design that will survive the 'fratricide' threat. A similar situation exists in the case of an attack on a reentering warhead by a hostile antiballistic missile (ABM) system. Here again, the computer is essential in designing a system whose vulnerability to an ABM attack is reduced to an acceptable level."

It should be no surprise to the Subcommittee that Russia has been seeking supercomputers for more than two decades. Investments of billions of rubles in trying to design their own supercomputers resulted in failure. Consequently, the Soviet government and then the Russian government sought to get such machines from the West, and pressed hard for disbanding COCOM in order to remove export restrictions.

China has gone down a similar path. Last year, when China carried out aggressive military exercises in the Taiwan strait, effectively closing the strait to both shipping and air traffic, the United States -sensing China might turn the exercise into a full scale invasion of Taiwan-moved two carrier task forces into the area. As the tension rose, a high ranking Chinese official threatened to launch nuclear ballistic missiles against Los Angeles. Such threats, and the willingness to make such threats, should make it clear that there are serious dangers today, and we should not want to exacerbate them by providing technology that will increase the risk and danger, as supercomputers will.

In light of the findings of the Energy Department, it is hard to grasp how the administration went down the path of making it easy to export and buy supercomputers. The result: for most transactions, the administration's supercomputer export controls are not any more burdensome than export controls on personal computers.

Today's export control regime on supercomputers is neither export controls nor a regime. Current regulations do not safeguard this technology and, in my opinion, are an exercise in Executive branch recklessness. I would strongly advise the Subcommittee and the Congress to legislate the immediate suspension of the supercomputer export regulations, because they are a clear and present danger to America's national security.

The current export controls on high performance (e.g., supercomputers) computers is found in the Federal Register of March 25, 1996. Put most simply, the regulation says that high performance computers can be exported without individual validated licenses, but there are some restrictions. Generally speaking, the restrictions apply to the country and end user the computers are going to.

The countries are organized into three groups or "tiers." Who is in each tier is, to a certain extent, bizarre and incomprehensible.

For example, the middle tier -countries that can receive supercomputers less than 10,000 Millions of Theoretical Operations Per Second (MTOPS) -includes Antiqua and Barbuda, Bangledesh, Belize, Equatorial Guinea, Haiti, Liberia, Nicaragua, Poland, the Slovak Republic, Somalia and Togo, as examples. Keep in mind that the entire Defense Department owns only two computers more powerful than these, and hardly any computers in this middle category.

Tier 3, which is a motley collection of countries as different as Angola, Belarus, India, Israel, Oman, Saudi Arabia and Tajikstan get computers in the range of 2,000 to 7,000 MTOPS.

Tier 1 are the most-favored countries, including our allies and a few others whose appearance on the Tier 1 list is hard to understand. For example, it includes Iceland, even though Iceland was never a COCOM member and never cooperated with the United States on export controls. The same holds for Liechtenstein and Luxembourg, places from which technology diversions were common in the 1970's and 1980's, San Marino is there too, but I don't know why. Tier 1 countries can receive any level of performance supercomputer. There are no restrictions on performance,

The caveats in the regulation relate to end use and end user, and the caveats are applied only where the end use or end user is nuclear, chemical, biological, or missile related.

On paper this sounds good, but in practice it is a meaningless restriction. Why? To begin with, the selling company has to "know" whether or not the "buyer" falls into one of these restrictive categories. But since there are no licenses and scant record keeping required, this restriction is hard to enforce. If anyone doubts this is so, consider the recent sales of supercomputers by Silicon Graphics that in 1996 went to a nuclear design installation in Russia, the Russian Scientific Research Institute for Technical Physics..

While this matter is still being under investigation, there can't be any doubt the computers are there in the service of Russia's nuclear weapons industry. This demonstrates clearly that the regulatory "regime" is a failure.

This is the first time any supercomputer been lost or gone to a nuclear weapons designer. It is a scandal and is destabilizing to world peace and security.

To make matters even worse, once a supercomputer is delivered it can be re-transferred and the U.S. government and the company are, in fact, out of the loop. For example, a supercomputer can be sold to a shoe maker in Iceland, and the shoemaker in Iceland can turn around and sell it to a Chinese missile factory. Because there is no longer an international licensing system or any other mechanism, it is reasonable to conclude that there is next to nothing we can do about such a re-export transaction.

There are other weaknesses in the regulations, but one that stands out is putting the burden of determining end use and end user on the exporting company. While I think companies must act reasonably and responsibly, end use and end users can only be ascertained by the U.S. Government, which has available to it extensive information developed through open and classified channels. Passing this responsibility off onto exporters is unfair and unwarranted, and it exposes our companies to penalties and potential criminal action if something goes wrong.

A particular problem, from my point of view, is the definition of supercomputers used by the administration. The threshold is set too high and the regulation does not take into account the ability to couple together certain types of supercomputers and near-supercomputers in order to enhance performance.

Today there are three major categories of supercomputers. There are Vector Processor machines (VECTOR), Massively Parallel Processor (MPP) supercomputers and Symmetric Multi-Processors (SMP). In addition there is the ability to cluster together high performance workstations, and different ways to link up and network the other types of computers, or drive them with different pre-processor engines.

The Symmetric Multi-Processor machines (examples include the Silicon Graphics Power Challenge and the DEC Alpha) can be linked together easily. According to an August, 1995 Defense Department study, SMP's "are liked by the DoD laboratories because of their associated lower costs and their architectural similarities with the vector computers .[and] some designs permit users to plug additional processors into the SMP factory enclosure up to some maximum number without factory support.. (emphasis added) An example is the Army Research Laboratory at Aberdeen Proving Ground, which has an 8-node (each with 12 processors) SGI Power Challenge computer. The DoD study says: "The system is currently treated as eight separate computers. Load sharing software is pending. This is the largest system SGI has fielded." The study notes: "ARL has a research agreement with SGI to develop Distributed Shared Memory and Distributed Virtual Memory software to implement cluster operations." This is the main resource of the laboratory and is shared among the Army (30%), Navy (30%), Air Force (30%), and other Defense agencies (10%).

The 2000 MTOP threshold put in place by the administration is questionable, since slightly lower performing machines can easily be linked together, as the administration's own study of High Performance Computing, Spring 1995, demonstrates. Many export control specialists, in and out of government, believe that dozens if not hundreds of machines have been shipped abroad. There is particular concern about China. Because of the limitations in current export regulations, there is no tracking on such exports.

There are those who say that, because the Defense Department has few supercomputers, this only shows it can do most of its work with the older machines it already has. This is not true. The Department clearly needs supercomputers, even in this era of restricted budgets; they will be the keystone for future defense systems which, more and more, will be based on high technology.

None of this stands in the way of those who want even more liberalization of export controls on supercomputers. This month a team, funded by the Commerce Department and (shamefully) by the Defense Department -the same team that prepared the arguments for massive liberalization of supercomputer exports, was back at the Pentagon working on yet another supercomputer liberalization. My sources say they are pushing to raise the threshold to 10,000 MTOPS, from the current starting point of 2,000 MTOPS!

The truth is that supercomputers are a critical tool for developing defense systems for the next century. Making such machines freely available to the world under the flawed system we now have, will help erode both our technology leadership and our national security. If the United States wants to retain its superiority in an era of collapsing defense budgets, it is critical to hold the line on these sensitive exports and keep these machines out of the hands of potential adversaries or proliferators. At the same time, we must make sure that the military departments and research activities of the Department of Defense have access to the best computing technology.

In summary, I suggest the following:

1. Suspend the current regulations on High Performance Computers.
2. Demand a full accounting of supercomputer sales under the current export regime.
3. Conduct a full assessment on the impact of computer sales on national security and on weapons proliferation.
4. Assess, using the CIA and Defense Intelligence Agency, who is seeking supercomputers and why they are wanted.
5. Develop and propose an effective export licensing system.

Disclosure Statement

Pursuant to Rule XI, Clause 2(g) of the Rules of the House of Representatives, this is to inform the Committee that I have no Federal grants or sub-grants and contracts or subcontracts with the Federal Government currently or for the past three years.

Dr. Stephen Bryen pioneered the field of technology security as Deputy Under Secretary of Defense from 1981-1988. Responsible for technology security policy and high tech trade matters affecting national defense, he worked to formulate national policies to protect U.S. military and commercial products, know-how, intellectual property, goods and services. Dr. Bryen founded the Defense Technology Security Administration (DTSA) and served as its first Director.

Dr. Bryen currently serves on the Government Security Committee of Space Systems/Loral in Palo Alto, California; the Boards of Directors of Telos Corporation in Herndon, Virginia and C-MAC Industries of Mechanicsburg, Pennsylvania. Dr. Bryen is a member of the Advisory Board of Beaconsfield Capital, L.L.C. (New York). He also is an Adjunct Fellow of the American Enterprise Institute and a Fellow at the Taiwan Institute for Political, Economic and Strategic Studies. In addition, he serves on the Advisory Board of the Jewish Institute for National Security Affairs. Dr. Bryen recently completed a special study on the "Revolution in Military Affairs and the Revolution in Technology Security" and a major project on "Cyber-terrorism -The Threat and the U.S. Response."

Dr. Bryen is also President of Delta Tech Inc., and a founder of the Intellectual Property Security Company. Delta Tech, a high technology/national security consulting firm, specializes in U.S. and foreign technology issues, promoting new technology, representing high tech companies, and identifying business opportunities for its clients. IPSC, with offices in the US and Israel, develops security applications for computers. Dr. Bryen is also senior Technology Adviser to Hollinger Digital Inc. in New York city.

In his dual capacities as Deputy Under Secretary of Defense for Trade Security Policy and Director of the Defense Technology Security Administration, Dr. Bryen and his staff of 130 were responsible for handling over 50,000 export license applications annually for goods and technology controlled under the Arms Export Control Act and the Export Administration Act. Dr. Bryen developed unique automation capabilities for his agency, reducing the time American business waited for a government export license ruling from an average of two months to an average of two days. Involved were licenses for armaments and armament technology and "dual use" technology such as semiconductors, computers, advanced electronics, machine tools, special materials, communications systems

In addition, Dr. Bryen helped enlarge and strengthen the Paris-based international Coordinating Committee for Multilateral Export Controls (COCOM). His work in modernizing the export control process led to the adoption at COCOM of a computer system designed by his agency, and similar systems in the COCOM countries was responsible for investigations of possible export violations and also for the creation of the structure of the COCOM control lists. Representatives from Dr. Bryen's office, and Dr. Bryen personally, handled international COCOM negotiations in Paris. Recently, the technology developed by DTSA has been made available to countries in Eastern Europe and FSU.

Dr. Bryen pioneered cooperative agreements with non-aligned and neutral countries such as Sweden, Switzerland, Finland, Austria, India, Korea and Singapore. In addition, Dr. Bryen -- who uncovered the "Toshiba case" diversion of sensitive technology to the Soviet Union -- worked with the Ministry of Trade and Industry in Japan in developing an effective export control program in that country.

SPECIAL EFFORTS ON THE SOVIET UNION, CHINA & IRAQ

During the 1980's a major focus of COCOM and DTSA was to prevent high level computers and communications equipment form going to the Soviet Union and other adversary nations. While preserving U.S. national security interests, Dr. Bryen also served as the Defense Department's negotiator with the Soviet Union on concluding a bilateral Space Cooperation Agreement. The agreement, made in 1988, facilitated space cooperation and remains in force today. A number of joint missions and programs, such as the Space Station, developed from this agreement.

During the late 1980's and early 1990's Dr. Bryen led the fight to prevent Iraq from acquiring weapons of mass destruction, such as nuclear, chemical and biological weapons and delivery systems for such weapons. In 1988, through his efforts, 1.5 million nerve gas antidote injectors were prevented from sale to Iraq. Iraq was then known to be producing nerve gas. In 1990, thanks to a tip, Dr. Bryen learned that Iraq was attempting to purchase special high temperature "skull" furnaces, ideal for nuclear warhead manufacturing. Building a coalition of government experts and Congressional supporters, the export was blocked by an Executive Order by President George Bush. In 1991 the Pentagon called upon Dr. Bryen to assess Iraqi technology acquisitions. Thousands of documents seized by U.S. Customs and the Justice Department were used to reconstruct the Iraqi acquisition network.

PRIOR POSITIONS

From 1971 until 1979 Dr. Bryen served as Foreign Policy Advisor to Senator Clifford P. Case (R-NJ) who was then Ranking Republican Member of the Senate Foreign Relations Committee and a senior member of the Senate Appropriations Committee. After 1975 Dr. Bryen served as a Professional Staff Member of the Senate Foreign Relations Committee and as the Staff Director of the Near East Subcommittee. He was responsible for foreign assistance legislation, for the State Department Authorization, for the Case Act (PL 92-403), for the Arms Export Control Act and for a host of legislative agreements. In connection with his work for the Committee, Dr. Bryen conducted study missions to Europe, Africa, Asia (including Vietnam and Cambodia), Rhodesia (now Zimbabwe), Cyprus and the Middle East as well as to the Soviet Union.

From 1967 to 1971 Dr. Bryen served as Assistant Professor of Government at Lehigh University (Bethlehem, Pennsylvania). He gave undergraduate and graduate courses in international politics and in methodology in political science. He is the author of The Application of Cybernetic Analysis to the Study of International Politics (Martinus Nijhof, The Hague, 1971) and (with Frank Colon) The Council of Governments System (State of Pennsylvania ).

EDUCATION

Dr. Bryen received is AB degree in History and Political Science from Rutgers University in 1964; his Master of Arts in International Politics from Tulane University (1966) and his Ph.D. in International Politics from Tulane University in 1971. His doctoral dissertation was: The Application of Cybernetic Analysis to the Study of International Politics. He was a General Electric Scholar at Rutgers University and a National Defense Scholar at Tulane University. He was sponsored for post-doctoral work in applied statistics by the National Science Foundation which was conducted at Virginia Polytechnic University.