I see a problem with this.
Were the models underlying those features trained on all available web content, or are they unlike any other enterprise models out there?
At any rate, you should see a bigger problem in what Google does, which you don't seem to.