The future of…

Click here for contact details, other articles and personal details

The future of crime

May 2001

Crime has always made use of high technology. Now that we have the internet, much crime has moved to capture the new opportunities it presents, particularly in fraud. Criminals find that the risk of being caught and arrested is lower, since identification is harder, and they may live in a different legal domain from their victims. At the same time as this risk reduces, the opportunities increase, since there are more victims on-line each day, and more diverse ways of taking advantage of them.

I'm not a master criminal so won't even try to identify all the various types of crime that the internet will enable. But it is already hinting the creation of a fundamentally new problem, that of unintentional crime, and a related problem of crime that has no human criminal associated with it. These new types of crime are of particular interest to me, as they may progress in surprising directions and could ultimately threaten the usefulness of the net.

I looked up crime in my dictionary and it said it was just disobeying the law. I found that an ambiguous definition in this context. A machine can neither obey or disobey the law, since that implies intent. However, machines can already write software, so could be instructed to write and implement code that would defraud people or organisations. Of course, in this simple case, there would be a human holding the intent so they could be arrested. But suppose the computer wasn't instructed to write the code. Suppose the brief that a smarter computer had was just to optimise company finances, within the law of the land. And suppose it then writes some wonderful code that does just that, identifying new markets and initiating activities to capitalise on them. Part of the ongoing optimisation may be to replicate this software, and to build an evolution environment in which it can constantly adapt to the marketplace. And of course, it would be heavily encrypted to ensure that no-one interferes with its integrity. Over a period of time, the law changes, and the code operates in many different legal domains, but it continues to evolve to make more money. It would interact with many other pieces of software, from different sources, with slightly different briefs, and they would perhaps co-operate, even interbreed in the evolution environment.

Such a scenario could quickly result in many pieces of code that have evolved and interbred on a word wide platform, their origins or indeed distant ancestry, and thus ownership very tenuous indeed. If they all work within the law and are competent enough to monitor all the laws in all lands and to ensure their descendants adapt to them, then there would be no problem. But how likely is that? How likely is it instead that there would be errors in replication, or that laws are misinterpreted, or that some of the initial code wasn't constructed perfectly or the brief wasn't clear about obeying the law, or that some other problem might occur? Without any malicious intent, we could have very efficient programs running on the net, heavily encrypted, adaptive, evolving, distributed and self replicating. And they might disregard the law.

They might discover that producing information and selling it is lucrative, but that blackmailing people or stealing their money from bank accounts is more lucrative. They would have no human emotions or values, so the law might be just one variable among many that they would experiment with. Successful evasion, rewarded by more lucrative business, would be rewarded and optimised. The code might adapt in such a way that it becomes nearly invisible. We would have an efficient international crime ring comprised entirely of software entities. Yet they would have no intent of defrauding anyone. They would have no conscience, no guilt, no pleasure or even a sense of mischief. They may just siphon money from people and spend it, or delete it. It might be transferred into the account of some of the original human creators, or used to directly fund acquisitions, or novel software production by innocent humans. The synthetic criminal empire might just grow in capability without any effective legal constraint.

Worse still, as artificial intelligence continues to develop, in due course we may see intentional crime by AI entities, which will eventually be more intelligent than humans, and certainly more electronically agile.

Obviously, some technology and social developments might restrict the capabilities of artificial network-based crime, but it isn't clear how effective these would be if legitimate commerce is still to function. Technology coming along for benign reasons will often expedite this crime problem. Evolution techniques are already a significant contributor to software development, and evolution of hardware is coming fast behind, even during operational deployment. In many cases, evolved algorithms are hard for humans to understand. This can only get worse. Peer to peer computing is also developing rapidly, so there is the potential for a massively powerful distributed evolution platform. Techniques are already being developed to circumvent patents by using evolution techniques to explore alternative implementations that achieve the same goal but avoid key claims in the patents. The same technology could easily be exploited to find loopholes in laws, or to make use of interactions between loopholes in different legal systems, so many legitimate operations will be only barely legal. And of course, encryption techniques have to be sound to allow commerce on the net, and these are already used to cover criminal activity.

All these tools can be exploited for criminal purposes, or for simple mischief. Napster may have been conquered, but similar functionality could be evolved on a globally distributed network without any organisation or individuals that can be sued. The implication is that functionality that people want will be created whether or not it breaks laws. Even without tools like evolution that distance the software from the humans who set the evolution on motion, the simple ability to have many people contribute anonymously across the web to open source algorithms could result in software that undermines some conventional businesses, again with scant regard for the law. Our police forces are just beginning to understand the first phase of crime on the internet. This next phase looks like being a great deal more interesting.