include "../includes/header.php"; ?>
This paper was originally written by David Forrest for Course TPP32: "Law, Technology, and Public Policy", Massachusetts Institute of Technology, Room 8-135, 77 Massachusetts Ave., Cambridge, MA 02139
The emergence of new technologies continually forces us to ask whether our laws provide the proper balance between protecting us from potentially harmful consequences of those technologies, and allowing us to reap the benefits. The development of nanotechnology, a molecular-precision manufacturing technology which is surprisingly close to realization, will seriously challenge the ability of our regulatory system to respond quickly and to maintain the critical balance between dangers and benefits.
The development of nanotechnology will affect regulations in most areas, from banking and commerce to air safety and toxic waste. This paper concentrates on regulations to assure that molecular machines and their products are developed in a safe and responsible manner. The aims of this paper are to:
Nanotechnology is an emerging manufacturing technology which will provide a very high level of control over the manipulation of matter [1, 2]. Machines, called assemblers, will be able to build things to atomic specification under programmable control. Their design and operation can be like robots and miniature factories, with levers, gears, bearings, electric motors, pulleys, cables, conveyer belts, and computers to coordinate their operationall with parts of molecular dimensions (Figure 1).
Because the laws that govern molecular systems are reasonably well-understood, the development of this technology does not depend on discovering new scientific principles. The advances we need are engineering advances: improved computational abilities (for the design of molecular systems) and improved capabilities for automated manipulation of individual atoms and molecules.
The development of nanotechnology is not dependent on advances in any single branch of technology. Any one or combination of advances in molecular chemistry, protein engineering, molecular biology, or scanning tunneling microscope technology would advance the development of nanotechnology. Regulation of nanotechnology should therefore cover all types of nanomachines along all the development paths. From what we know now, we can project at least three distinct ways to develop an assembler with the ability to construct arbitrary structures:
Each of these paths could proceed in one of two ways: (1) first build limited assemblers which could build more sophisticated assemblers which would build even more sophisticated assemblers, and so on, until a general assembler is built, and (2) directly build an assembler with general capabilities if the manipulation tools are adequate.
At first glance, the ability to put atoms and molecules exactly where you want them may not seem so spectacular an achievement. We already can make things to micron-scale precision, so why should an extra three decimal places make that much of a difference? The answer lies principally in this: with nanotechnology you can manipulate any kind of atom or molecule, not just the highly specialized materials of microelectronics or biology. And with this general capability to make structures, it will be possible to make assembler systems which can be programmed to make copies of themselves. After the relatively short time it takes to make a trillion trillion assemblers, they can be reprogrammed to make other things rapidly and cheaply. The following examples illustrate some of the consequences of general assembler capability.
Materials. With the ability to build things atom-by-atom, nanomachines will be able to build any physically possible machine or structure, given enough time, space, matter, energy, and design capability. Estimates show that, with nanotechnology, materials such as steel could be made ten times stronger and more ductile than is possible today. Materials made from diamond could be even stronger than that, and three times lighter. These supermaterials will allow dramatic improvements in the performance of machines limited today by material capabilities; applications include power generators, automobiles, jets, and spacecraft.
Computers and Artificial Intelligence . Molecular computers a trillion times smaller than today's machines and at least as fast as current devices will be possible. Molecular tape memory, similar in principle to the way information is stored in DNA molecules, will allow storage densities of one gigabyte per cubic micron. This means that we could store all of the information contained in the Library of Congress in a cube the thickness of a sheet of paper. To process this information, calculations show that we could have a desktop-sized computer which contains the raw processing power of one million human brains.1
Medicine. Molecular computers will be small enough to fit the equivalent of a mainframe computer within the volume of a typical human cell. These computers could direct machines to repair DNA and other damaged cell structures and destroy cancerous cells, viruses, and bacteria. Death due to disease and aging would be eliminated [4, 5].
Environment. Self-replicating assemblers could scour toxic and nuclear waste sites, transforming carcinogens into harmless chemicals and separating radioactive isotopes. Materials which people now discard as trash could be recycled within the household to make food, clothing, and household goods. The acts of dumping, pouring, and exhausting chemical by-products into the environment will become practices of a bygone era. Nanotechnology will allow us to make things without dangerous by-products; leftover atoms can be recycled or made into harmless chemicals.
Economy. Nanotechnology will alter the economy of the world, changing our concepts of what things are of value. Ordinary dirt will become a useful construction material with nanotechnology, and therefore more valuable. Diamond will become a common, inexpensive substance, useful as a construction material for nanomachines and household items. Food will be virtually free, because the machines to make it will be self-replicating and the materials are abundant in dirt, water, and air. Transportation of goods will be mostly unnecessary; the information needed to make a desired product can be sent electronically to a local assembler system.
Military Systems. Applications to offensive, defensive, and military surveillance systems abound. The same could be said for law enforcement systems, terrorist weapons, and devices for criminal purposes. Examples include virus-like machines able to identify people by their genetic code and attack or spy on them, and plagues of omnivorous "locusts" able to ravage continents.
1 In terms of gate operations per second. This does not automatically imply one million times the intelligence of a human. That will depend on the capabilities of future software.
From the general implications given above, we can see that nanotechnology will radically alter several aspects of the way we live and work. Along with the many changes, we will have to insure that we can still enforce our existing laws as well as develop new laws to keep pace with new problems and challenges that nanotech-nology will bring. This is the task of our policymakers, and it will require a substantial effort given the range of issues cited in the previous section.
This technology is very close to realization, but how close? Prediction of the timing of future developments is a notoriously difficult task, made even worse by the multi-disciplinary nature of nanotechnology. The only certainty is that the developments will occur. Factors that will influence the speed of development are:
It is not unreasonable to expect a U.S. initiative on nanotechnology, once politicians become aware of its importance.2 This will accelerate development. Current estimates (by those familiar with the issues) of when an assembler will be developed range between the years 2000 and 2030 [6, 7].
It is worth considering the possibility of simply banning nanotechnology altogether, in which case we would avoid the need to expend further effort on technology policy questions. There will be enormous pressures to develop nanotechnology because of the potential benefits to society, as well as the threat of others gaining military superiority should they develop it first. It does not seem likely that all nations would agree to ban development. Even if they did, verification would be nearly impossible because research efforts could be easily hidden in small laboratories. And because of the multi-disciplinary nature of nanotechnology, one would have to ban a large fraction of scientific research because so many areas will impact on developing nanotechnology, e.g.scanning tunneling microscopes or computational chemistry. So it seems unlikely that we could implement a verifiable worldwide ban.
If we tried to block or slow the development of nanotechnology in the United States, or in other democracies, we would increase the chances that nanotechnology is first developed in a country without a free press. In which case we could not be certain that that country would not use nanotechnology to oppress its neighbors or the rest of the world. So efforts to slow progress only serve to threaten our own freedom.
Therefore, a sensible course of action when formulating nanotechnology policy is to assume that nanotechnology will be here sooner than most people expect (the ten-year time horizon) and concentrate on guiding development to avoid the dangers instead of blindly opposing development. When we consider what must be accomplished in that time frame, it seems clear that we should begin the task as soon as possible.
2 Efforts related to nanotechnology, such as the Kunitake Molecular Architecture Project and the Frontier Research Program, have already started in Japan.
It is necessary to make some assumptions regarding the course of development of a future technology in order to make some reasonable statements about the regulation of that technology. A major assumption that I make is that the U.S. will pursue the following policy in developing nanotechnology: (1) concurrent with developing an assembler, develop a "sealed assembler laboratory"  for safe experimentation with molecular assemblers (the lab would allow only information in or out, not atoms); (2) develop an "active shield" to protect the biosphere against hostile nanomachines.3
On the basis of these assumptions, part 4 concentrates on issues relating to the safe development of nanotechnology. Instead of exploring the full range of all possible regulatory frameworks or alternatives to regulation, I took the approach of presenting one plausible scenario and, in doing so, highlighted issues which would also be important in other approaches to regulation. A regulatory agency which could assume the responsibilities described does not currently exist; presumably, the cleanest way to implement this proposal would be to write new legislation and create a new agency, clearly defining any areas of overlap with existing regulatory agencies.
3 The active shield, as described by Drexler, would be a cornerstone of defense policy. The shield would be an automated system because the time scale of attacks could be on the order of minutes or secondsfaster than ordinary human response times. It would protect the environment against hostile systems regardless of their origin . For example, it could protect Soviet, Chinese, American, and European territories equally against Soviet, Chinese, American, and European nanomachines. This would therefore be a stabilizing defense initiative.
Common law and the U.S. Code are the rules to which we ideally adhere in our society. One of the most difficult far-term problems that nanotechnology will bring is how to enforce those rules. Whether current penalties will be stiff enough to deter certain crimes, and whether we will even be able to catch criminals at all, are likely to be areas of much debate. Central to this debate will be the issue of what level of surveillance is necessary to enforce the lawthe classic conflict between (a) the right to privacy and personal freedom and (b) the need for public safety.
One unknown is the amount of surveillance on human activities which would be necessary for effective protection. People may decide they want a high level of surveillance if it would eliminate violent crimes. The amount of freedom a society must sacrifice depends on the range of options our technology can provide (still largely unexplored) and the level of risk that that society, and other affected societies, are willing to accept. While our attention will be mostly devoted toward shorter-term regulatory problems, we must not forget to think of them in the context of this most important overall goal of preserving and extending the freedoms we now enjoy.
Slightly modifying the language of the Toxic Substances Control Act (TSCA) provides a reasonable starting point: we should regulate when a device based on assembler technology whose manufacture, processing, distribution in commerce, use, or disposal may present an unreasonable risk of injury to health or the environment. The term "manufacture" should include making the device under laboratory conditions as part of the development effort. Devices which would potentially fall under this category include: (1) machines based on assembler technology able to autonomously make significant modifications to the natural environment outside of the machine,4 (2) machines that can self-replicate in the natural environment,5 accounting for the full range of possible environmental conditions and for materials normally encountered in the environment. Machines, or their decomposition products, which have chemically toxic properties would be covered under TSCA and would probably not need additional legislation.
It is important to recognize that, while the release of a replicating assembler into the natural environment is an area for concern, there is little danger of any resulting problem unless the device was intentionally designed to operate in that environment especially if the device is designed in accordance with a set of safety standards. The primary concern is the theft of an assembler by those who would abuse its capabilities. Thus, both regulatory issues and national security issues must be considered.
The next question involves the notification mechanism by which a regulatory agency knows that someone is developing a machine with the above properties and should begin monitoring for compliance with applicable regulations. Again, using TSCA as a model, anyone wishing to manufacture a new chemical must give prior notice to the EPA for review under the pre-manufacturing notice requirement . In the face of civil and criminal penalties (as in TSCA) most firms and researchers would comply with the notification requirements. Laws could also restrict companies from doing research in countries which do not have similar regulations. Apart from that, national security concerns may dictate the need for a task force to actively seek out those who might develop such machines in secret; the regulatory agency would probably not be directly responsible for this type of activity.
4 The wording is intended to cover, for example, steel-eating termites.
5 The term "environment" includes the insides of people and other life forms.
In this assumed scenario of nanotechnology development there are four distinct phases for which different amounts of regulatory control are required. This suggests a broad outline for phase-in and phase-out of regulatory control during the most dangerous periods of development.
|Phase 2Post-assembler, Pre-assembler lab|
|Phase 3Post assembler lab, Pre-active shield|
|Phase 4Post active shield|
Phase 1 is where we are nowassemblers have not yet been built. A regulatory framework should be developed in parallel with the development of assembler technology. Careful thought and critical debate about the problems now will avoid needless fears about the safety of this technology. The federal government could act without necessarily regulating by promoting and assisting the writing of standards. As developers get closer to the assembler "breakthrough," security precautions should increase. Regulations and an operative regulatory agency should be in place before the breakthrough.
In Phase 2, assemblers have been successfully developed, but no assembler-proof containment facilities are available. This situation requires close monitoring for compliance with standards, but even more importantly, assemblers (and the techniques used to make them) must be protected against theft by those who would abuse the technology. An important goal during this phase is to develop sealed assembler labs so that widespread experimentation can be performed without the need for security measures. Advantages of careful inspections are reductions in developers' proof-of-compliance paperwork, and enhanced assurance to the public that development efforts are conducted in a safe manner.
Avenues for commercial development could be made available during Phase 2, with extensive safety reviews of both the manufacturing process and the product by an appropriate regulatory agency. This would provide a near-term economic incentive for developers to advance the technology, reduce the government's funding burden, and provide the public with some near-term benefits of the technology. Here, a sorting process would be helpful. Products that do not contain replicators or assemblers6 could be deferred to existing regulatory agencies for review. Products that contain non-replicating assemblers7 would undergo thorough review by an agency with appropriate expertise in assembler technology. Products and manufacturing devices that contain replicating assemblers would undergo the most stringent levels of review; quite possibly, they should not be made available until they can be extensively tested under safer conditions during Phase 3. A further distinction could be made depending on the evolutionary capacities of the machines. Although the only systems described so far would not have the ability to mutate and evolve, it is conceivable that such a system could be designed at some point.8 Machines in this category are of special concern since their functions and capabilities could change over time. Sorting should be performed by an agency familiar with all aspects of assembler technology and be performed according to due process procedures.
In Phase 3, sealed assembler labs have been developed, which means that experiments on assemblers can be safely performed by anyone. If these labs are made widely available to people wishing to develop advanced assemblers or other devices and materials, there will be little incentive for these people to develop their own assemblers outside of assembler labs. That is, provided that they have some outlet for turning their designs into safe products. Assemblers in a sealed box are fine for experiments and development work, but not much good if you want to get a finished product out. A logical approach would be to have the same review channels as described for Phase 2 commercial development. Once a person's design clears these channels, the product could be manufactured by a producer with an "open" assembler operating under heavy security.
In Phase 4, the active shield has been developed and is in place. Assemblers outside of sealed labs could be made widely available. Many regulatory controls (e.g., the close monitoring of Phase 2) can be phased out and regulatory alternatives can take over, because bad nanomachines can be contained. It is difficult to say what regulatory alternatives will be most effective, because that will depend on what levels of potential damage and surveillance the society has decided it is willing to tolerate.
For example, the tort system may or may not be as effective as it now is. With molecular assemblers to cheaply repair injuries and damages, any liability claims are likely to be quite small. Since the liability system has not historically remedied minor problems and risks , and most problems will be minor because of nanotechnology, this mechanism may fade into disuse. However, if the court system also takes advantage of this new technology, it may be possible for a person to collect damages quite easily; with electronic transactions (and yes, maybe electronic lawyers and judges) a successful lawsuit could be accomplished in minutes from one's living room.
Market forces are likely to be much more powerful. With rapid access to accurate information on how good a product is, and who likes it and who doesn't, consumers can rapidly select against poorly-designed products and for the safer, more reliable, and otherwise most appealing products. But this will depend on the extent to which people will value safety and reliability, and that is currently unclear. Product features other than safety and reliability may have much greater perceived value.
6 For example, pocket supercomputers.
7 For example, a machine that makes only food. These non-reprogrammable limited assemblers are devices into which you dump a bucket of raw material and out of which comes a bunch of widgets. The widgets, of course, should be safe products as determined by CPSC or other applicable institutions.
8 Such a system would have to be designed; reference  presents specific arguments why evolving nanomechanical systems would not occur by accident.
Apart from intentional abuses, there are many reasons to believe that systems made using nanotechnology can be extremely safe. First, there are the market forces just mentioned. Second, because of nanomachines' high degree of control over matter, dangerous manufacturing by-products can be avoided.9 And Drexler has already proposed several good solutions to the problem of nanomachines accidentally replicating out of control.10 These approaches to safety could be incorporated into a set of standards. Regulatory enforcement of, or voluntary compliance with these standards would be desirable to ensure safety. But questions remain: What institutions will develop these standards? How will they develop them? Will the standards be enforced, and if so, by whom? In addition, we may want to avoid having enforcement procedures significantly slow down or interfere with legitimate development efforts.
Developing a set of safety standards for experimental procedures, a set of standard tests to determine when an assembler may be hazardous, and a set of standard assembler designs will be an important part of the proposed regulatory framework. One type of standard now in use is a consensus standard. Voluntary consensus standards are developed by organizations with the participation of interested partiesproducers, users, and general interest groups. Due process safeguards are incorporated into the rules for developing these standards, allowing for airing of diverse viewpoints and means of appeal [15, 16]. Industry compliance is voluntary, unless the standards are subsequently adopted by a regulatory agency. The technical expertise available in the private sector makes this a valuable approach for developing safety standards, as pointed out by Baram :
The advantage of utilizing the private sector's technical expertise in formulating health, safety, and environmental regulatory standards cannot be overemphasized. It is a fact that this expertise cannot be matched, in the vast majority of instances, by the technical staffs of federal, state, and local regulatory authorities. In addition, the utilization of active technical standards-writing committees from the private sector is an efficient and dependable means of ensuring that standards are kept up to date with developing technology.
Existing standards organizations, such as the American National Standards Institute (ANSI) and the American Society for Testing and Materials (ASTM), are well-equipped to do this procedurally, but currently do not have expertise in nanotechnology. The same holds true for testing laboratories such as Underwriters Laboratories, Inc. As nanotechnology develops, these institutions should be encouraged to develop the expertise needed to formulate a sound set of standards.
Questions regarding the appropriateness of voluntary consensus standards still remain: How can we assure ourselves that due process will be followed in developing these standards? Even if it is followed, will the standards be good enough? Will everyone voluntarily comply with all of the standards? If they don't, what are the chances of a serious accident? And even if all this is ok, it may look bad politically if the government isn't directly involved.
Baram  points out that, historically, success with using nongovernmental standards as an alternative to regulation depended on two conditions: (1) the technologies and risks were well-understood, and (2) potential liability was significant enough to force responsible industry behavior. The potential liability of a runaway replicating assembler is the worth of our biosphere, price enough to insure significant caution. But nanotechnology may not be sufficiently well-understood to merit this voluntary approach. Furthermore, most sources agree that if the potential effects of the substance or product in question are clearly irreversible or hazardous to human health or the environment, that item should be subjected to standards enforcement [19, 20, 21]. Some products of nanotechnology could fall into that category. This is the primary argument for regulatory control of nanotechnology development efforts, and why alternatives to regulation would be inappropriate.11
Recognizing that although there is a chance that it will not be necessary to regulate the development of nanotechnology, for the purposes of this paper I assume that the arguments are compelling enough that it will be regulated.
9 Note that dangerous by-products will still be possible, e.g., due to poor nanomachine design or by malicious intent. The difference is that, unlike with current technology, the cost of eliminating hazardous waste will be small.
10 A potential danger, that of a nanomachine escaping from a lab and replicating trillions of copies of itself at the expense of the environment, can be neutralized by a number of engineering solutions: limiting the number of copies a machine can make of itself; making machines that need special, rare molecules in order to operate; and making machines that can work only in special laboratory or industrial environments .
11 For example, if we used strict liability as an alternative to regulation it would be impossible for any developer to internalize the cost of the risk (destruction of the biosphere), so theoretically the activity of developing nanotechnology should never be undertaken . Besides, if civilization is destroyed there won't be anyone around to collect damages.
Government regulation can provide an extra measure of security that voluntary standards do not: a mechanism for enforcement. Enforcement involves monitoring for compliance and taking action when standards are violated. The standards developed by other organizations could be adopted by a regulatory agency with the power to enforce them. Adopting these standards would also be accomplished via a due process mechanism which would serve as a backup to the one used to develop the standards.
But regulatory control has its share of problems as well:
A variety of solutions to these problems have been proposed; the best should be sorted out and incorporated into future nanotechnology legislation. For example, for Item (7), the American Bar Association proposed that regulatory agencies establish policy consultation boards to help consider broad policy issues . For Item (3) Drexler has proposed that a hypertext publishing system would be an effective way to incorporate a much wider cross-section of the public's knowledge and ideas into the decision-making process [34, 35].12 For Item (6), when writing new legislation for regulation of nanotechnology, the authors could clearly delineate the respective roles of any new regulatory agency and existing agencies.
For Items (5) and (8), one way to reduce the compliance paperwork burden on the developers (and their researchers) would be to place on-site inspectors at the most critical development centers, and assign roving inspectors to secondary sites during Phase 2. The inspectors would be experts in appropriate disciplines, and, to reduce chances of corruption, would work in teams, be well-paid, and periodically rotated. There could be due-process-compatible mechanisms for rapid incorporation of overlooked problems into the regulations, based on inspectors' observations. If the developers themselves have a reasonable amount of input into the standards to which they must comply there is likely to be an inherently good level of cooperation with inspectors. This proposal pre-supposes that there will either be a small number of critical development centers, requiring a modest number of inspectors, or that the government would be willing to pay for a large number of inspectors because of the high risks involved. If the development centers are for-profit institutions, a special tax could be assessed to share the burden of inspections. This level of scrutiny would only be necessary until Phase 3, after which there would be little need to experiment outside sealed labs.
It therefore seems that a reasonable framework for assuring safe development of nanomachines is a standards-enforcement approach, where all interested parties would have a say in developing standards and experimental practices via due-process safeguards. Enforcement would be the responsibility of a regulatory agency backed by the soundest legislative provisions we can devise, written in clear language after considerable thought and debate.
12 Although this system is still under development, indications are that it will be implemented in 1989 or 1990giving us sufficient lead time to evaluate the concept for future legislation. [Editor's note: This time frame estimate turned out to have been rather optimistic. For current information on Foresight's hypertext publishing projects, see http://www.foresight.org/WebEnhance/index.html]
Before active shields are developed, standard design elements of nanomachines, standard experimental procedures, and performance standards can be useful guides to avoid replication accidents or other unwanted effects. The former two are specification standards, which require a particular implementation ('you must limit replicating ability using a counter limit embedded in hardware,' or 'assemblers with replication ability must be kept in a Level 3 containment facility'). They have the advantage of being easy to monitor for compliance, but limit "the range of technological change"  which will be especially needed during the development process. Performance standards require achievement of some measurable result ("you must limit the replicating ability of an assembler to 20 generations"). They allow more flexibility for implementation, but require more monitoring to assure compliance. During the early development stages performance standards and standards for experimental procedures (specifically regarding containment) would make sense. Once a variety of designs have been tried and tested, the most successful could be converted into specification standards, preferably with the option of superseding them when even better designs and procedures become available.13
After the development effort, the active shield will be in place to prevent outbreaks of errant and hostile nanomachines, so why should the latter have to conform to any safety standards? Unless the implementation of the active shield is extremely fine-grained (e.g., several defensive machines for every cubic micron of matter on the planet), there could be local damage by bad nanomachines. If the active shield is fairly coarse-grained, there might also be significant costs associated with fighting back large hoards of bad nanomachines.14 In determining who should bear the burden of those costs, it would be necessary to learn whether the developer was negligent. The existence of industry or government standards and practices, and whether the developer adhered to them, would be useful criteria.
Verification of the presence of a standard design in a questionable nanomachine might also be a useful tool for the active shield. One of the active shield's criteria for suppressing certain machines or research efforts might be violation of these standards.
13 I have in mind a standard design for a component which limits the replication of an assembler and which could be incorporated into a variety of assembler designs, in the same way that a flywheel governor is a limiting device which works on a general class of machines: rotating engines.
14 As the shield goes to finer and finer grain, shield costs go up, but defense and repair are cheap because you neutralize the bad machines before they replicate too much. As the shield graininess coarsens, shield costs go down but defense costs and amount of local damage increase. The optimum balance will depend on factors such as how much potential damage a society is willing to risk, the complexity of the shield, and the amount and degree of intelligent surveillance a society is willing to tolerate.
In describing Phase 2 and 3 commercial development I alluded to a system in which approved designs could be turned into products through a limited number of highly-secure facilities. Without restrictions on the facility owners, they would have an instant monopoly on virtually every product because it will be much cheaper to make most things using replicating assemblers than by using any existing manufacturing process. What would be needed is not only a review process for product safety, but ways to provide fair access to the facilities for both individual citizens and industry.
It seems possible to do this by treating these facilities as regulated public utilities, with due process mechanisms for both individuals and companies to gain access. Users could pay the costs of producing a product, including the cost of the safety reviews. Conforming to standard designs which have already passed prior reviews would keep costs down and allow quick turnaround through the safety reviews. Non-profit organizations, venture capitalists, and other institutions could sponsor designs of exceptional merit to help developers who cannot cover the costs alone.
A major problem with this scenario is still how to decide, from a large set of competing products, which get priority at these facilities. Which will benefit society most? Economic analysis may allow us to optimize the use of these limited resources, except in cases where the benefits cannot be expressed in economic terms (e.g., number of lives saved). It will not be necessary to tie up the facility to make multiple copies of a product, since the facility could make one copy of a limited assembler which could in turn make unlimited quantities of the product that you want. So limited assemblers will nearly always be more cost effective than products which cannot make other things.
Whether or not these facilities should be for-profit entities is debatable. For example, there are instances of nuclear power plants ranging from unsafe and inefficient to both safe and efficient. In the final analysis, it is most important that the open assembler facilities be safe from two perspectivesmilitarily secure from outside access, and internally secure so that only authorized products get out. Since the facilities would eventually be phased out once the active shield is implemented, efficiency of operation would not be of long-term significance.
Although this paper concentrates on issues related to regulating the safe development of nanotechnology, other laws will have to be created or modified because of the technological advances possible with replicating assemblers. Following are some representative issues that have not been addressed by existing laws and which highlight the importance of more research on nanotechnology policy.
Machines' rights. With nanotechnology a brute force approach to artificial intelligence would be to map the architecture of a person's brain, atom for atom, and simulate its operation on much faster nanocomputers. If this machine is effectively smarter than a human, do we have the right to turn it off? To own it? What criteria can we use to know what rights a machine should have? How much decision-making control should machines have?
Rights to our uniqueness. With nanotechnology it will be possible to copy a person, atom-for-atom. Will this be legal? Can a government make copies of its best soldiers, even with their consent? What if you copy yourself, but your wife doesn't want to do the same? Who is married to her?
Death. Large numbers of people will die between now and the development of cell repair machines. As we approach that development point, it will become increasingly evident that we can suspend the decay of people just about to die (by freezing them), wait until we get cell repair technology, and then make necessary repairs and start them up again . In today's society we do just about everything possible to keep people alive. Do we therefore decide that it is society's responsibility to begin placing dying people in suspension? When do we start, since cryonic technology is available now?
Distribution of land and wealth. At the point when everyone can own replicating assemblers we will have a situation where a few individuals own large amounts of land (read "land" as "suddenly useful raw material for making lots of useful things") and lots of people owning some or none. Will this mean that wealth is radically redistributed the instant that cheap replicating assemblers become available to convert the previously low-value dirt into high-value product? Or will replicating assemblers command such a high price that only wealthy land-owners can afford them?
The incentives to develop assembler-based systems for defense purposes will be enormous, so we can expect a substantial initiative. There are a variety of foreign policy and strategic issues here which fall beyond the scope of this paperhopefully the initiative will be directed towards more stable options like an active shield, than toward offensive weapons. One area which is relevant to the subject of this paper is the safety of classified research. Unfortunately, the number of options available to insure that classified research is conducted safely seems limited. Two options are as follows:
- The government initiative can generate its own set of safety standards and practices, and enforce them internally. If the government initiative is created by a law, the legislative provisions could explicitly state procedures for generating and enforcing safety standards, although the prospects for enforcing the law would be lower without direct public scrutiny. Congressional oversight may help.
- A regulatory agency or commission, separate from the defense initiative, could oversee the development of internal standards (making them consistent with any relevant unclassified standards), and monitor the initiative for compliance with those standards. The amount of due process possible in this type of situation is necessarily limited, since relatively few numbers of people would have access to all of the information needed for critical debate. It may, however, be possible to select a set of people with a diversity of viewpoints (via normal due process channels) and give them the necessary security clearances for the task.
Although option (b) represents unprecedented scrutiny of defense activities, careful selection of those who would have access to classified information would help insure its secrecy. The knowledge that the safety of classified research is being independently monitored will help assuage public fears, even though they cannot gain access to the classified information. The motivation for a separate agency to oversee safety compliance stems from the federal government's historic failures in self-regulation: e.g., in nuclear power plants  and the Department of Defense's Biological Defense Research Program .
Nanotechnology policy is an important area for research. We should aim to have most of the regulatory framework laid out within the next ten years, before assemblers with general capabilities to manipulate matter are developed.
There are potential dangers, such as runaway replication, which can be avoided by design. It seems that regulatory control will be necessary to assure that nanotechnology is developed safely. Safe designs, safe experimental procedures, and methods to test for potentially hazardous assemblers can be incorporated into standards by consensus of interested parties. The standards can be adopted and enforced by a regulatory agency. The greatest danger appears to be intentional abuse of the technology, so certain aspects of development should be performed in a secure environment.
A plausible scenario is to phase in regulatory controls during the most dangerous times, and then phase them out when they become unnecessary. Four specific phases have been identified:
Phase 1Pre-assembler: Safety standards are developed in parallel with the technology; regulatory controls are phased-in; security is increased to protect key aspects of assembler technology and key personnel.
Phase 2Post-assembler, Pre-assembler lab: Close monitoring of key developers by a regulatory agency to assure compliance with safety standards; heavy security of facilities with "open" assemblers; efforts are directed to develop sealed assembler labs; commercial products can be made available after review by a regulatory agency.
Phase 3Post assembler lab, Pre-active shield: Most research can be done safely in sealed assembler labs, so most close monitoring of Phase 2 is no longer necessary; still need heavy security of facilities with "open" assemblers; commercial products can be developed by anyone and submitted for review prior to manufacture by a secure facility; other efforts are devoted toward developing an active shield.
Phase 4Post active shield: Environment is safe from assemblers, secure facilities are no longer needed; regulatory controls exist as necessary.
Development for defense purposes is likely and there is reason for concern of inadequate compliance with appropriate safety standards. A regulatory agency may be able to monitor the defense initiative for compliance while satisfying both the initiative's need for secrecy and the public's need for assurance of safe research practices.
The concepts in section 3 (guiding nanotechnology development rather than trying to block or slow development) were originally advanced by Eric Drexler, Foresight Institute. The ideas in section 4.6 were discussed during a weekend retreat of the MIT Nanotechnology Study Group, 5-7 April 1985, Groton, New Hampshire, and at subsequent meetings.
 K. Eric Drexler, Engines of Creation, Doubleday, New York (1986).
 K. Eric Drexler, "Molecular Engineering: an approach to the development of general capabilities for molecular manipulation." Proceedings of the National Academy of Sciences (USA), v. 78, pp. 5275-5278 (September 1981).
 K. Eric Drexler, Transcript of a talk on Nanocomputers, presented at Xerox Palo Alto Research Center, California (22 May 1986).
 K. Eric Drexler, "A Technology of Tiny Things: Nanotechnics and Civilization," Whole Earth Review (Spring 1987).
 A. K. Dewdney, "Nanotechnology: wherein molecular computers control tiny circulatory submarines," Scientific American, v. 258, no. 1, pp. 100-103 (January 1988).
 K. Eric Drexler, panel discussion in the symposium "Nanotechnology: Prospects for Molecular Engineering" (12 January 1989).
 Discussions at various MIT Nanotechnology Study Group meetings (1985-1989).
 K. Eric Drexler, Engines of Creation, Doubleday, New York, pp. 184-186 (1986).
 K. Eric Drexler, Engines of Creation, Doubleday, New York, p. 198 (1986).
 K. Eric Drexler, "Biological and Nanomechanical Systems: Contrasts in Evolutionary Capacity," in: Artificial Life, SFI Studies in the Sciences of Complexity, ed. C. Langton, Addison-Wesley Publishing Company, p. 16 (1988).
 Toxic Substances Control Act, Section 5, "Manufacturing and processing notices," United States Code Annotated, Title 15, p. 412.
 K. Eric Drexler, "Biological and Nanomechanical Systems: Contrasts in Evolutionary Capacity," in: Artificial Life, SFI Studies in the Sciences of Complexity, ed. C. Langton, Addison-Wesley Publishing Company (1988).
 Michael S. Baram, Alternatives to Regulation, D.C. Heath and Company, Lexington, MA, p. 14 (1982). KF5407 .B37
 K. Eric Drexler, Engines of Creation, Doubleday, New York, p. 183 (1986).
 Michael S. Baram, Alternatives to Regulation, D.C. Heath and Company, Lexington, MA, pp. 53-54 (1982).
 1980 Annual Book of ASTM Standards, Part 10, American Society for Testing and Materials, Philadelphia, PA, pp. iii-iv.
 Michael S. Baram, Alternatives to Regulation, D.C. Heath and Company, Lexington, MA, pp. 60-61 (1982).
 Michael S. Baram, Alternatives to Regulation, D.C. Heath and Company, Lexington, MA, p. 56 (1982).
 United States Senate, Committee on Governmental Affairs, Study on Federal Regulation, Vol. VI, "Framework for Regulation," U.S. Government Printing Office, Washington, p. 276 (1978). KF5406 .A25
 Michael S. Baram, Alternatives to Regulation, D.C. Heath and Company, Lexington, MA, p. 153 (1982).
 American Bar Association, Commission on Law and the Economy, Federal Regulation: Roads to Reform, American Bar Association, Washington, DC, p. 9 (1979). KF1609 .A74
 Michael S. Baram, Alternatives to Regulation, D.C. Heath and Company, Lexington, MA, p. 25 (1982).
 United States Senate, Committee on Governmental Affairs, Study on Federal Regulation, Vol. I, "The Regulatory Appointments Process," U.S. Government Printing Office, Washington, p. xxxi (1977). KF5406 .A25
 United States Senate, Committee on Governmental Affairs, Study on Federal Regulation, Vol. III, "Public Participation in Regulatory Agency Proceedings," U.S. Government Printing Office, Washington, p. vii (1977). KF5406 .A25
 Roger G. Noll, Reforming Regulation, The Brookings Institution, Washington, DC, pp. 5-9 (1971). JK901 .N793
 American Bar Association, Commission on Law and the Economy, Federal Regulation: Roads to Reform, American Bar Association, Washington, DC, p. 92 (1979). KF1609 .A74
 United States Senate, Committee on Governmental Affairs, Study on Federal Regulation, Vol. IV, "Delay in the Regulatory Process," U.S. Government Printing Office, Washington, p. ix (1977). KF5406 .A25
 United States Senate, Committee on Governmental Affairs, Study on Federal Regulation, Vol. V, "Regulatory Organization," U.S. Government Printing Office, Washington, pp. xv-xvii (1977). KF5406 .A25
 American Bar Association, Commission on Law and the Economy, Federal Regulation: Roads to Reform, American Bar Association, Washington, DC, p. 70-72 (1979). KF1609 .A74
 American Bar Association, Commission on Law and the Economy, Federal Regulation: Roads to Reform, American Bar Association, Washington, DC, p. 100 (1979).
 American Bar Association, Commission on Law and the Economy, Federal Regulation: Roads to Reform, American Bar Association, Washington, DC, p. 70 (1979).
 George C. Eads and Michael Fix, Relief or Reform?, The Urban Institute Press, Washington, DC, p. 192 (1984). HD3616 .U47 .E23
 American Bar Association, Commission on Law and the Economy, Federal Regulation: Roads to Reform, American Bar Association, Washington, DC, pp. 100-101 (1979). KF1609 .A74
 K. Eric Drexler, Engines of Creation, Doubleday, New York, p. 226 (1986).
 K. Eric Drexler, "Hypertext Publishing and the Evolution of Knowledge," Draft 2, unpublished (December 1987).
 K. Eric Drexler, Engines of Creation, Doubleday, New York, pp. 133-139 (1986).
 "Study Raises New Doubts About DOE N-Plant Safety," Engineering Times, p. 12 (November 1988).
 Melissa Hendricks, "Germ Wars," Science News, v. 134, no. 25, p. 392 (17 December 1988).
 K. Eric Drexler, "Nanomachinery: Atomically Precise Gears and Bearings," in: Proceedings of the IEEE Micro Robots and Teleoperators Workshop, Hyannis Cape Cod, November 1987 (in press).include "../includes/footer.php"; ?>