I just went through this discussion in an EU funded project, and the consensus was that complying is up to the individual partners. In our case all academic partners decided to comply by posting preprints on their websites (contentwise identical to the accepted version), which is allowed by both ACM and IEEE.
I really dislike this approach. Finding the free version is harder and the links are much less likely to be permanent. I've found a huge amount of data, code and extra material is simply missing now. In 20 years are you confident someone will maintain the same url and content?
Can you afford the open access option with the journals you want to submit to?
