Research and Advances
Computing Applications Digital rights management and fair use by design

A Skeptical View of Drm and Fair use

Don't expect DRM to ever be smart enough to distinguish fair use from copyright infringement.
Posted
  1. Introduction
  2. Implementing General Fair Use
  3. Approximations
  4. Special Cases
  5. Conclusion
  6. Author

Digital rights management (DRM) technologies allow users to view and manage content in a controlled fashion; for example, a DRM system may allow a piece of recorded music to be played but not copied. >>>> All of the various types of DRM systems operate by restraining a work with some kind of technological lockbox; the work might, for example, be protected by encryption. Any use of the work requires the participation of some special hardware or software acting as a gatekeeper to determine which uses proceed and which are blocked. This gatekeeper mechanism can implement any policy it likes for determining which accesses to allow.

There are, however, very good reasons to doubt whether DRM can ever be effective. There is no theoretical basis for believing that DRM systems can withstand attacks, and in practice DRM has had a checkered history. Although the general prospects for DRM deserve skeptical discussion, that is not my topic here. I assume the use of an effective DRM system in position to prevent or allow each attempted use of the copyrighted work.

U.S. copyright laws give copyright owners the right to prohibit others from copying a work or creating a derivative work. Fair use can be understood as an exception to this rule, saying that in certain cases a user can legally copy a work or make a derivative work, even if the copyright owner objects; for example, including short clips from a movie in an online review of the movie is generally fair use, even though it involves copying parts of the original copyrighted work. Similarly, making a backup copy of a digital work is typically fair use.

The legal definition of fair use is, by computer scientists’ standards, maddeningly vague. No enumeration of fair uses is provided. There is not even a precise algorithm for deciding whether a particular use is fair. Instead, the law says that judges should make case-by-case decisions based on four factors: the nature of the use; the nature of the original work; the portion of the original work used; and the effect of the use on the market. The law does not say exactly how these factors should be evaluated or even how the factors should be weighted against one another.

To a computer scientist, such imprecision is a bug; to lawyers it is a feature, since it allows judges to take into account the unique circumstances of each case. Making fair use a judgment call allows the fair use doctrine to evolve in light of technological innovation. It provides a kind of flexibility and adaptability that would not be possible with a more precisely specified rule.


An approach that makes errors in only one direction simply makes too many errors, so we must accept that any practical system is both too permissive and too restrictive.


The lack of a bright-line rule separating fair use from copyright infringement has another consequence. Although some uses are clearly fair and others clearly not fair, there is a large gray area of uses that may or may not be fair. Even a well-trained copyright lawyer cannot say with certainty where the line lies between fair and unfair uses. The most anyone can say about one of these gray-area uses is that there is a certain probability that a judge will find it legal.

Back to Top

Implementing General Fair Use

The vagueness of the fair use test makes it essentially impossible to create a DRM system that allows all fair uses. To be correct, such a system would have to apply the four-factor fair use test to each attempted use of a work. The nature of the test makes it impossible for two reasons:

Lack of knowledge about the circumstances. Aspects of the test require knowledge about the circumstances of the use, but such knowledge is not available to the DRM system; for example, a certain use may be fair when done in a classroom but illegal when done in a commercial setting. The DRM system cannot know enough about the circumstances outside the computer to know whether the setting would more accurately be classified as teaching or as commerce.

Inadequate artificial intelligence. Even if full information about the circumstances were available, applying the four-factor fair use test would require highly sophisticated AI. Several of the factors involve "AI-hard problems." For instance, the fourth factor in the test evaluates the effect of the use on the market for the original work. It requires reasoning about the economics of a particular market, a task even well-trained humans find difficult. For the foreseeable future, no computer system will be able to approach a human’s ability to analyze these markets.

A DRM system that gets all fair use judgments right would in effect be a "judge on a chip" predicting with high accuracy how a real judge would decide a lawsuit challenging a particular use. Clearly, this is infeasible with today’s technology.

Back to Top

Approximations

If our technologies can’t make the fair use judgment correctly in every case, perhaps they can get it right most of the time. Perhaps they can enforce some approximation of the law.

The challenge in doing this lies again in the difficulty of internalizing the four-factor fair use test so a program can evaluate it. The true result of the test relies on economic analysis and on factors outside the computer that are not easily measured (such as the social context in which a use occurs). In some respects, the fair use test seems designed to frustrate attempts to computerize it.

It is difficult to imagine how to do an approximate economic analysis or how to approximate the social context of a use. In practice, an approximate algorithm would have to ignore these factors or replace them with crude proxies. For example, if an accurate fair use analysis would look favorably on classroom uses, an approximation algorithm might be more permissive when running on computers registered as belonging to schools or that are running the "education version" of some software product.

A plausible approximation algorithm would make errors in both directions, allowing some uses the law would forbid and forbidding some uses the law would allow. Consider, for example, the system’s evaluation of attempts to copy an entire copyrighted work. There are situations in which this is fair use, as well as many in which it is not. The system often cannot tell them apart. So if the requirements say the system must prevent all unfair uses, then apparently it must flatly refuse requests to copy the entire work—and thereby ban backup copies. Alternatively, if the requirements say the system must allow all fair uses, then apparently it must allow virtually all requests to copy the entire work—and thereby allow blatant infringement. An approach that makes errors in only one direction simply makes too many errors, so we must accept that any practical system is both too permissive and too restrictive.

Such a system would have drawbacks for both copyright owners and legitimate users. Overly permissive cases provide loopholes for people who want to infringe copyrights, so copyright owners would suffer. Overly restrictive cases would prevent legitimate users from making legal and expected uses.

Back to Top

Special Cases

If DRM systems can’t make the right judgment in every case, perhaps they can get some special cases right. Perhaps they can allow backup copies or personal use within the home. Perhaps these special cases are simple enough that they can be reasonably approximated.

But even these seemingly simple cases are more difficult than they might initially seem. A backup, for instance, is most useful if it can be restored on a different machine (in case the original machine breaks). But backup cannot simply provide a mechanism for moving a file from one machine to another; such a general file-transfer facility is a ready loophole for infringers. The solution may involve centralized record keeping, ensuring a backup is not restored too frequently or in too many places, though such record keeping raises privacy issues.

The point is not that handling backup is impossible but that it is surprisingly challenging. To date there has been no satisfactory solution to these problems, though it may be because most of the development effort has been (mis)directed toward the effort to build all-encompassing DRM systems. There may be hope, however, for a bottom-up approach that tries to handle a few cases well.

Back to Top

Conclusion

Fair use is one of the starkest examples of the mismatch between what the law requires and what technology can do. Accurate, technological enforcement of the law of fair use is far beyond today’s state of the art and may well remain so permanently. Technology will not obviate the need for legal enforcement of the copyright rights of both copyright owners and users.

Back to Top

Join the Discussion (0)

Become a Member or Sign In to Post a Comment

The Latest from CACM

Shape the Future of Computing

ACM encourages its members to take a direct hand in shaping the future of the association. There are more ways than ever to get involved.

Get Involved

Communications of the ACM (CACM) is now a fully Open Access publication.

By opening CACM to the world, we hope to increase engagement among the broader computer science community and encourage non-members to discover the rich resources ACM has to offer.

Learn More