This is not a good analogy. An academic environment has different strictures and norms than a business environment.

In a business environment, the relevant questions are:

1) does it break any laws? 2) is it profitable? 2b) is it consistent with the image MS wishes to project to its customers and/or regulatory overseers

One person's unfair competition is another's brilliant hack.

My comment wasn't meant to reference whether it was wrong or right, just the logistics of what is happening.

Think of it this way: Your professor asks you for the name of the imaginary friend that the professor's son always talks about. You have no idea what the answer could possibly be, so in your head you choose "John," a reasonably common name with a very low probability of being right. Google answers with a strong conviction in it's voice that the answer is "Mark," another common name. You have no evidence to believe that your own guess is correct, so in the face of the appearance of belief on Google's behalf, aren't you the least bit tempted to say that it's Mark and not John?

Customers care about results, not about how hard you worked to get those results.

This is a fundamental difference between student reward-schedules and rest-of-life reward schedules.

There's no credit for effort, unless you can spin it that way in your marketing / branding.

Horrible analogy. This isn't a test. This is a product.

Actually it is a test, and how well you answer it is directly responsible to how your business will or won't succeed.

Google has worked out their algorithms for processing the incoming data and generating an answer. Bing has apparently used Google's answer as a comparison to whether they are getting it right or not and when not, subbing in the other answer.

While in business this may not be illegal, is it still very much 'Not knowing the correct answer to the test'.

No, it's a product. Come on. Every product is a "test" of the market so in that one definition of "test", yes. But not in an academic sense, which is what you meant. Because only in an academic sense does the concept of "cheating" exist.

Look at it like a machine learning problem. If you slavishly match the test set you are doing something called overfitting. Your performance on the rest of the web will decrease, because you inherit bias from the test set.