The basic questions posed are 1) would the special safeguards proposed by
the British be effective in reducing the risk of misuse of the computers
to an acceptably low level, and 2) would approval seriously erode
existing computer export controls or could this application be
considered as constituting a unique case? The Committee could not agree
on answers to these questions and the different views are given in the
enclosed report.
The Committee was also unable to agree on recommendations with respect to
the policy options. DOD, JCS, AEC, and STR recommend that we reaffirm our objection to the sale.
Commerce favors reaffirmation of our objection; but, if your decision is
to approve the transaction, believes that approval should be predicated
upon prior acceptance by the three governments concerned of very
explicitly articulated safeguards and procedures for implementing them.
State, Treasury, OST, CEA, and USIA recommend that we lift our objection on condition that
the UK agree to effective implementation of the proposed safeguards and
to support continuation of tight controls on computers and technology at
the next COCOM list review.
Enclosure4
Washington, March 16, 1971.
NSC UNDER SECRETARIES COMMITTEE
REPORT
SUBJECT
- British Proposal to Sell Two Large Computers to the
USSR
I. Introduction
In response to Mr. Kissinger’s
memorandum of January 25, 1971,5 the Under
Secretaries Committee has reviewed an outstanding U.S. objection in
COCOM to the export by a
United Kingdom firm (ICL) of two large computers to the Institute of
High Energy Physics at Serpukhov in the USSR. Prime Minister
Heath raised this matter
during his visit with the President and with Secretary Rogers. In order to assist the
Committee review, a technical study on this matter was prepared by a
panel convened by the Office of Science and Technology (OST) and is attached to this
report.
The Institute at Serpukhov conducts unclassified basic research and
engages in extensive exchanges of data, publications, and personnel
in international cooperation efforts with institutions and
scientists from free world countries, including the United States.
The principal instrument around which the Institute’s activities
center is the world’s currently most powerful (76 Gev) proton
accelerator.6 An appropriate computer
facility has been designed to complement the research potential of
the Institute. The computers proposed for export exceed the
performance standards of the BESM 6, which is the best Soviet
computer available to the Institute, and will satisfy only
approximately one-half of the Institute’s foreseeable computational
requirements.
When the U.K. presented this proposal to COCOM in October 1970, we objected to it because of the
risk of diversionary use of the computer facility for other than its
intended purposes as well as the precedent-making implications which
the export could have for controls over advanced computers.
Additionally, the Joint Committee on Atomic Energy had urged that
advanced computers not be sent to the USSR.
[Page 933]
Because of the unanimity rule in COCOM, and the inclusion of more than $3 million in
U.S. parts and components, the British Government has been
constrained from licensing the proposed export although the U.S. is
alone in objecting. Italy and the Netherlands had reserved judgment
at the time the U.S. raised its COCOM objection in October, 1970.
A refusal by the United States to approve the case will offend the
British, for whom this $11 million sale is a significant transaction
in terms of their hope to be a principal supplier of large computers
to the USSR. It also may disappoint some members of the U.S.
scientific community whose work involves them with the Institute at
Serpukhov.
In order to overcome the U.S. objection, the British Government, in
cooperation with ICL, has since outlined a series of safeguards
which they believe will reduce the risks of misuse of the computers.
Safeguards include (1) the planning of as full a research schedule
as possible for the machines, (2) the necessity of written
supporting documentation for individual program runs, (3)
contractual rights of free access by specialists of free world
parties, which have cooperation agreements with Serpukhov (presently
U.S., U.K. and France), (4) 10-year control over spares and on-site
maintenance, and (5) ICL willingness to obtain Soviet agreement
before export of the computers to permit U.K. personnel to empty
memory cores on demand of their stored informational contents and to
transmit them for U.K. (and U.S.) governmental analysis.
II. Security Aspects
No transfer of production technology would be involved in this case.
The security risk is the potential clandestine use of the computers’
capacity for strategic military purposes, e.g., advanced weapons
design computations. The risk of diversion depends on the answers to
two questions:
- 1)
- Can the proposed safeguards be implemented?
- 2)
- If implemented, will the safeguards be effective in
reducing the risks of misuse of the computers?
There is general agreement that the answer to the first question is,
for the most part, yes. On the second question, opinions differ.
State believes that the safeguards bring the risk of misuse to an
acceptably low level. DOD believes
that the safeguards are ineffective since they offer no high
probability that diversion, if attempted, would be eventually
discovered.
Basis for position held by State
[Page 934]
The OST panel study deals with the
British presentations on safeguards as well as the technical
feasibility of establishing absolute safeguards and detection
against misuse of the facility.
The OST panel believes that diversion
of less than of the order of 25 percent of capacity for two or more
years would not be worth the effort required of the Soviets to
effect clandestine diversion. The panel believes that this much
diversion of the computers’ time from legitimate work needs would
likely be sensed by foreign specialists working at the Institute and
who will be familiar with its research programs. The study concludes
that while complete elimination of risk of clandestine misuse is not
possible, the risk of such diversion is low. The study posits that
it is not possible to assert with certainty that diversion, if
attempted, would be discovered. However, the study notes that the
ability to “dump” the contents of the computers exposes the Soviets
to a finite danger of discovery provided these dumps are analyzed by
U.S. personnel. Experts believe that someone familiar with high
energy physics programs and weapons programs would have a good
chance of detecting a significant illicit program in the core,
provided he had a knowledge of the machine’s language and an
awareness of the program supposedly being run at the time of the
dump.
State accepts the technical views of the panel, noting that it
consisted of distinguished scientists with extensive experience in
the fields of computers, high energy physics and strategic weapons
design. State suggests, moreover, that because of the presence of
foreign scientists at the Institute and the core dump provision, the
Soviets would not consider the Institution’s computers to be fully
secure for handling classified programs despite the machines’
location on Soviet territory. Hence, they would be extremely
unlikely to use weapons programs on such systems. The Soviets would
be much more likely to use an additional Soviet-built BESM 6, which
has approximately 25 percent of the power of the total British
system, for weapons calculations than to try to divert time from an
installation not completely under their control.
State notes that the British have supplied information subsequent to
the OST report which demonstrates that the main and secondary
storage capacity planned for Serpukhov is similar to that provided
for comparable U.S. and U.K. installations.
Basis for position held by DOD
The fundamental flaw in the safeguard proposals is that when the
concrete steps required to put them into meaningful practice are
examined, the potential effectiveness of the safeguards appears
increasingly remote.
For example, the first three safeguards proposed by the British
assume that discrepancies between computer usage and the needs of
legitimate programs would be quickly noted. Yet, according to the
OST study, a knowledgeable U.S.
scientist working at Serpukhov and knowing the various experimental
and test programs there could assess the
[Page 935]
computational load only within a factor of
two. In other words, so far as these three safeguards are concerned,
a diversion of approximately half of the computational capacity
might go undetected. To achieve even this much control would require
considerable and continuing experience with the installation. Thus,
these safeguards depend on the presence at Serpukhov of
knowledgeable U.S. scientists willing to carry out a monitoring
scheme and able to remain at the installation for extended tours of
duty. DOD believes such a U.S.
presence would be required on all shifts.7
The OST report lists the “dumping” of
the computer core as the only special safeguard offered by the U.K.
which has any teeth against a determined effort at diversion. In
this connection, it is worth noting that it is only the content of
the cores, e.g., the internal memory of each central processing
unit, and not all stored informational content which is subject to
dumping on demand.
Thus, the OST panel stated it could
not assert that diversion, if attempted, would be eventually
discovered with a high degree of probability. It pointed out that
schemes are possible to “capture” the executive system and replace
it with one which “looks” the same externally on legitimate work but
which allows “hidden” programs to be run. This means that an illicit
program might be detected only if it happened to be in the central
processing unit when the dump was made. It is at this point that the
OST panel’s questions as to why
the system includes unusually large external memories become
especially relevant.
Assuming the right to demand core dumping is exercised at
sufficiently frequent intervals to capture any illicit programs,
there is, according to a report of the National Academy of
Sciences,8 no presently
developed methodology for analyzing the contents of such a dump.
Moreover, the National Academy of Sciences has estimated that an
expenditure of several hundred thousand dollars of research effort
over a period of one to three years would be needed to develop such
a methodology.
[Page 936]
In addition, DOD believes it would be
necessary to create in the U.S. a group of high-energy physicists,
computer center managers and weapons designers to analyze the
recorded data obtained from each dump. How large this group would
need to be and whether it would be engaged on a full or only
part-time basis will depend upon the frequency of the dumps and the
methodology developed for analyzing their contents. Unless this
aspect of the problem is addressed and solved, including the funding
of such a group, even the core dump provision will be an empty
safeguard.
The 10-year control over spares and on-site maintenance is less a
safeguard against diversion than the basis on which sanctions would
be applied if a diversion were detected. It is, of course, a threat
whose impact will depend on whether the USSR has any reason to
suppose the U.K. would actually cut off support if a diversion was
suspected and proved. In this connection, it is not easy to imagine
such a drastic sanction being employed on the basis of the kind of
evidence which the other safeguards are apt to provide.
It is worth noting that when the Soviets two years ago expressed
interest in a comparable U.S. computing system, the CDC 6600 for
Serpukhov, an interdepartmental review, backed by two outside
studies, concluded that the U.S. Government would be unwilling to
issue an export license except possibly under safeguard conditions
which would be either unacceptable to the Soviets or too expensive
to carry out on the U.S. side.
Comments by other Agencies
AEC believes that the OST panel based
its conclusion of low risk of diversion largely on U.S. presence at
Serpukhov, although in the AEC view there is waning interest in
Serpukhov and there may not be any U.S. scientists there after the
present program is completed later this year.
On the other hand, OST points out
that the panel’s estimates were based on Western presence, not just
U.S., and that the British have agreed to obtain random core dump
printouts. OST believes that there
will be continuing U.S. interest in the Serpukhov accelerator even
after the large U.S. accelerator at Batavia becomes available.
III. Precedent-Making Implications
A difference of Committee views also developed in assessing the
precedent-making implications of a decision to lift the U.S.
objection to the export of these ICL computers.
Position of State
Regarding the potential impact of this export on COCOM and U.S. export controls, the
international stature and importance of the Institute
[Page 937]
as the largest in its
field, its willingness to accept extraordinary safeguard conditions
and its openness to foreign scientists, can be considered as
constituting a unique situation which justifies a single exception.
Should the British export be approved, possible proposals by Control
Data Corporation to sell large computers to other Soviet research
institutes such as those at Yerevan and Dubna, as well as any
subsequent proposals, would necessarily be subjected to detailed
investigation on their individual merits in accordance with the
standards applied in this case. An additional factor relating to
large CDC machines, which does not apply in this case, is that there
are weapons codes compiled in CDC computer machine language.
Pressures for relaxation of computer parameters which exist within
COCOM are endemic and can be
anticipated to continue regardless of the outcome of the British
proposal. As far is known British ICL is the only non-U.S. company
presently seeking orders in Eastern Europe for computers which
exceed the COCOM technical limits.
A further consideration is that the Soviet agreement to on-site
inspections by Americans on their territory could, if carried out,
establish a useful precedent which might facilitate future arms
control negotiations. Finally, the safeguards in this case bring it
within the U.S./U.K. guidelines for dealing with computer exceptions
requests.
Position of DOD
There is little doubt that this transaction, if approved, would
create an undesirable precedent for both COCOM and U.S. controls. The OST panel noted that each of the two systems exceeds
the COCOM guidelines in every
relevant specification by a considerable margin. To approve this
export and not do great violence to COCOM controls, it would be necessary for the U.S. to
claim that the Institute is a uniquely deserving end-user, that its
personnel are not likely to be required by the Soviet Government to
divert the system and that, in any case, the safeguards included in
the contract effectively eliminate the risk of such diversion. If
these claims are made, the predictable result would be the prompt
submission, by U.S. firms as well as others, of a number of requests
for the export of very large computers to other destinations in the
USSR. Each request will claim the end-user to be deserving; each
will assert the improbability of diversion; and each will propose to
include safeguards similar to those accepted for Serpukhov. It will
be difficult for the U.S. to object to these requests since by
approving the Serpukhov case, it will have, in effect, certified the
adequacy of the Serpukhov safeguards. If the safeguards were, in
fact, adequate, this result would be entirely acceptable. The
crucial fact, however, is they are not, and to the extent that they
can be used to justify computer exports the net result will be the
rapid destruction of existing U.S. and COCOM controls.
[Page 938]
The British could go ahead despite lack of COCOM approval, but would then have to consider the
need for U.S. unilateral approval to ship the estimated $3 million
of U.S. parts and components which are included in their proposed
export.
Comments by other Agencies
The AEC believes that a favorable U.S. decision could result in the
establishment of undesirable precedents in our international embargo
control scheme as well as increased pressures from both U.S. and
foreign computer firms to sell embargoed computers to the bloc. It
would be extremely difficult to maintain present U.S. and
international controls if such a major exception is made to these
embargo controls. The U.S. has been the primary spokesman among our
Western European allies over the years for restraining trade with
the bloc in large, strategically-useful computers. A major departure
from this position, as would be represented by approval of the U.K.
case, would erode significantly our posture on computer controls
that the U.S. has long held with its Western allies. AEC believes
that U.S. companies, particularly CDC, will push for approval of
several large computer exports to the Soviet Union if the U.K. case
is approved.
OST believes that the security risk
from this one transaction is not significant. Approval should be
predicated on the uniqueness of the Serpukhov facility, its use and
needs, as well as a clear understanding with the British that it
would not represent a precedent for frequent exceptions of such
magnitude. It notes that there are at present no other 76 GEV
accelerators in the USSR or the world, which could similarly qualify
as unique international institutions.
Policy Options
1. Reaffirm U.S. objection in COCOM.
PRO
- a.
- Would avoid incurring any security risk.
- b.
- Would avoid the precedent-making implications which would
flow from approval of computers of this size.
- c.
- Would permit additional time to develop more effective
safeguards as well as to determine if safeguards negotiated
with the U.K. in previous cases are effective.
CON
- a.
- Prime Minister Heath’s
raising this subject with the President underscores the great
importance the British attach to it. Reaffirmation of our
negative position after their extraordinary efforts to satisfy
our concerns would be very disappointing to the British.
- b.
- The Soviets would cite our decision as further evidence of
U.S. reluctance to permit peaceful trade in goods containing
advanced technology.
- c.
- If this exception is not approved, the U.K., supported by
others, can be expected to press for relaxation of COCOM controls during the coming
List Review.
2. Lift our objection.
PRO
- a.
- Would be responsive to Prime Minister Heath’s personal approach to
the President.
- b.
- The on-site inspection arrangements could establish a possibly
useful precedent for future arms control negotiations.
- c.
- Would provide increased opportunities for Western cooperation
in high energy physics at Serpukhov and might cause some general
improvement in the Soviet attitude toward cooperative scientific
ventures.
- d.
- Would ease pressures for sales of computer production
technology to Communist countries by demonstrating that
occasional sales of advanced hardware can be approved.
CON
- a.
- Would provide the USSR with an advanced computer system better
than any they now have and capable of being diverted to
important strategic uses. Until additional details are developed
on implementation of proposed safeguards, it is not possible to
establish that such diversion would be detected.
- b.
- Would encourage U.S. and other firms to attempt sales of large
computers to the USSR and Eastern Europe.
- c.
- Would cause increased pressure by American manufacturers for
relaxation of U.S. export controls.
- d.
- Would encourage some COCOM
members to demand a more flexible policy toward exceptional
exports of other strategic items and technology.
- e.
- Would set the stage for demand by other COCOM members for a major
reduction in computer embargo levels at the List Review.
3. Lift our objection on conditions that 1) U.K. agree to effective
implementation of the proposed safeguards, particularly joint
procedures for the inspection and random core dump provisions
including the transmission of the core printouts to the U.S., 2)
U.K. agree to support continuation of tight controls on computers
and technology at the next List Review.
[Page 940]
PRO
- a.
- Would be responsive to Prime Minister Heath’s personal approach to
the President provided agreement can be reached on effective
implementation of the safeguards.
- b.
- Would involve the British with us jointly in a safeguards
implementation procedure, which could provide valuable
experience to both countries and have important implications for
future transactions in the computer field.
- c.
- Would tend to reduce the unfavorable precedent-making
aspects.
- d.
- Would provide increased opportunities for Western cooperation
in high energy physics at Serpukhov and might cause some general
improvement in the Soviet attitude toward cooperative scientific
ventures.
- e.
- Might establish a precedent for inspection or safeguards of
some future value in arms control negotiations.
CON
- a.
- Could be expensive in U.S. manpower and money to implement
safeguards.
- b.
- Would delay execution of ICL contract until implementing
details were worked out.
- c.
- Would require USSR to accept controls possibly beyond those
already contemplated.
- d.
- Would probably still entail some security risk.
- e.
- Might encourage U.S. and other firms to attempt sales of large
computers to the USSR and Eastern Europe.
- f.
- Might encourage other COCOM
partners to demand equal treatment on other strategic items and
technology.
- g.
- Export of these large computers by the U.K. would tend to
reduce the value of U.K. support for continued tight controls
over computers in COCOM.
DOD, JCS, AEC and STR
recommend Option 1. State, Treasury, OST, CEA and USIA recommend Option 3. Commerce
favors reaffirmation of the objection as in Option 1; but, if the
President’s decision is to approve the transaction, believes that
approval should be predicated upon prior acceptance by the three
governments concerned of very explicitly articulated safeguards and
procedures for implementing them.
Approve:
Option 1—reaffirm our objection
Option 2—lift our objection
[Page 941]
Option 3—lift our objection on conditions that 1) UK
agree to effective implementation of the proposed safeguards,
particularly joint procedures for the inspection and random core
dump provisions including the transmission of the core printout to
the U.S., 2) UK agree to support continuation of tight controls on
computers and technology at next list review9
