First, let's analyze the laws one by one:
1. A software robot may not injure a human being or, through inaction, allow a human being to come to harm.
There are some ways a web application can 'injure' a human being. For example, exposing someones private data to wrong people can cause all sorts of trouble ranging all the way from being dumped because of some photos from a party to having your house robbed while you are on a vacation by burglars who had your address and read your status message 'We are on vacation!'. Various ways of hurting people or allowing them to get hurt may include child pornography, cyber stalking (possibly in combination with real stalking) allowing them to get robbed, etc.
2. A software robot must obey any orders within its scope given to it by authorized human beings, except where such orders would conflict with the First Law.
This one is pretty straightforward, since it encompasses a requirement common to all software - authentication and authorization. I would also put user friendliness and overall user experience in this category because it allows an easier way for humans to communicate their orders to software. Maybe it's a stretch, but even performance metrics like response time could fit nicely here.
3. A software robot must protect its own existence as long as such protection does not conflict with the First or Second Law.
As I said in my previous post, this one should be easiest to implement, since it basically implies that software should be stable enough not to crash or delete its own data.
As you may have noticed, software requirements imposed by this interpretation of Asimov's laws are not in any way new. Respect of privacy, user friendliness, fast response time and overall stability have always been important aspects of software design. What is important about Laws of Robotics isn't in any one particular law - it's in their order of importance, and that is the main point of this article.
In order to be Asimov compliant, not only does software have to meet all these requirements, but also in correct order.
Example 1: Let's say that some hacker tries to steal personal data from an online service. Our software detects unauthorized intrusion, however has no way to stop it other than to shut down the server or otherwise cause downtime. Now, 'Asimovian' software has to protect its own existence in order to comply with the Third law. However, the First law says that human beings may be harmed if their personal information leaks to shady people. The First law is more important than the Third law, and software has to shut down and cause downtime, rather than to expose its users to risk.
Problems with such approach are quite evident:
1. Downtime = loss of money.
2. If a company shuts down their server to prevent information theft, they are practically admitting they have been hacked. Much better to just pretend it didn't happen and that no one would find out anyway.
Guess what, people generally do find out when their credit card gets drained. And they hate it much, much more than when they see a 404 page!