Basically, they are taking a rather low-key approach: the weakest form of hybrid open-access, where the journal and author retain full rights to the material, the default is subscription-only access, with authors allowed to post the articles on their own or their institution’s web site, but with authors allowed to pay ($3000) for open access.
This is weaker than the model I favor, which is that used by the journal Bioinformatics, published by Oxford University Press. It differs in one particular: the Bioinformatics articles become open access after a year, while the IEEE articles remain hidden behind a pay wall forever if the author can’t pay the open-access fee. I suppose that the difference is that IEEE believes that there is little time value for their articles, and that no one would pay for a subscription if they could just wait a year to get the material for free. Given the slowness of publication in computer science and electrical engineering, there may be some truth to this belief, as anyone who needs timely access can’t rely on IEEE journals for their information anyway.
I think that the IEEE may be a bit too cautious, but I’m more comfortable with that than with the eager embrace of open-access journals (without much thought about how unfunded researchers will get published) of the ISCB, whose policy I posted about in ISCB open access policies. Perhaps I’d feel differently if I were a generously funded researcher who had never had a dry spell in funding.