Here's an interesting and informative article on SearchEngineWatch about how pretty much all of the big search-engine companies give advance information on work in progress, if you know where to look. But one thing listed here doesn't meet the sniff test.
Check out this page on Google's web site, entitled "Papers written by Googlers." If you read carefully, you discover that what they are really saying is that they've hired people who used to write real research papers. Google (according to their web site) was founded in September of 1998. By definition, anything before that was not done by Google employees (with the very notable exception of Page and Brin's papers on PageRank, giving credit where it's due). Of the remaining papers, almost all of them were published by current employees before they joined Google. Of the small handful that actually list a Google affiliation, most are people who joined Google between when they did the research and when it was officially published.
Google is trading on their employees' past work and reputations. The company contributed nothing, other than to hire them and lock them away where they can no longer publish. From what I've heard, they pay their employees well and I'm sure the stock options for those employees are great compensation. But they shouldn't pose as an active collaborator and supporter of the open computer science research community, when it's clear they're really not. Here's the real list of papers from Google employees, all eight of them.
That's right, eight. By contrast, here's the IBM Research list, the Microsoft Research list, and the Yahoo Research list. Sun. HP. Intel. Nokia.
Once again, I respect their decision not to let their people publish research papers, i.e. not to participate in the larger research community and advance the field. I professionally disagree, but I respect that they made a business decision and are staying the course. What I object to is Google giving the false impression, for PR and recruiting purposes, that they are an active participant and collaborator, when they clearly aren't.
OK, I'm done ranting. Don't forget to read the whole SearchEngineWatch article -- it's good reading.