On balance, the use of encryption, just like the use of good locks on doors, has the net effect of preventing a lot more crime than it might assist.

As the local police department might want to decrypt a phone of a criminal suspect, so would the Chinese or the Russian or the Iranian intelligence agencies like to be able to do exactly the same thing.

We basically have only two real tried and true techniques that can help counter this. One of them is to make systems as simple as we can, and there are limits to that because we can only simplify things so much. The other is the use of encryption.

As we build systems that are more and more complex, we make more and more subtle but very high-impact mistakes. As we use computers for more things and as we build more complex systems, this problem of unreliability and insecurity is actually getting worse, with no real sign of abating anytime soon.

If we try to prohibit encryption or discourage it or make it more difficult to use, we're going to suffer the consequences that will be far reaching and very difficult to reverse, and we seem to have realized that in the wake of the September 11th attacks. To the extent there is any reason to be hopeful, perhaps that's where we'll end up here.

What encryption lets us do is say, Yes, the Internet is insecure. Bad guys are able to compromise computers everywhere, but we're able to tolerate that because if they do intercept our messages, they can't do any harm with it.

The perspective that law enforcement is presenting seems to be a very narrow one that's focused very, very heavily on investigations of past crimes rather than on preventing future crimes. It's very important for policymakers to take that broader view because they're the ones who are trusted to look at the big picture.

In order for any smartphone manufacturer to decrypt the data on your phone, it has to hold onto a secret that lets it get that access. And that secret or that database of secrets becomes an extremely valuable and useful target for intelligence agencies.

Computer science doesn't know how to build complex systems that work reliably. This has been a well-understood problem since the very beginning of programmable computers.