Rowan Collins wrote:
These are probably hosted on the Wikimedia Commons (which is a repository of liberally-licensed multimedia content,
No, it's not. It is a repository of liberally-licensed multimedia content plus, paradoxically, a few restrictively-licensed logos from the supposedly-so-free Wikimedia Foundation.
if you do get the Commons dump, you can be fairly sure there are no images in it that are actually copyrighted,
The majority of content in the Commons is copyrighted. If it wasn't, it couldn't be licensed, liberally or otherwise.
Timwi