Very probably yes. GCHQ's retention goes back (very broadly) that far (it's inconsistent) - and they have some access to each other's networks (the NSA access G

Very probably yes. GCHQ's retention goes back (very broadly) that far (it's inconsistent) - and they have some access to each other's networks (the NSA access GCHQ's network via example.gchq.nsa.ic.gov, specifically, so for example https://wiki.gchq.nsa.ic.gov/index.php https://wiki.gchq.nsa.ic.gov/index.php would get you to GCHQ's internal MediaWiki if you're on the NSA network - GCHQ use an internal .gchq TLD). They can query selectors on each other's databases, and filtering of what's "not allowed" seems to be very often up to the analyst. (Of course, I don't know if they share everything. Probably not.) The NSA's own BLARNEY initiative for this stuff dates back to 1978, although I think GCHQ beat them to the punch in effective mass telephony intercept - the first one I'm aware of started in the early 1980s, although that is offline, and probably no longer archived as the media (digital tape with low-bitrate encoded audio) would probably have degraded beyond all usefulness by now (and it sounded pretty crunchy in the first place, I understand, kinda like a bad squelchy lower-side-band transmission!). Remember, however, that they didn't have the same kind of analysis capability back then that they've got these days (but they can probably go back and analyse old stuff they still retain). As Snowden's disclosed, you're seeing the newer systems having full-take ring buffers in nearline storage, followed by offline selection for recent access and analysis from that using a huge amount of distributed processing, and in turn automatic selected archiving out of that. It's pretty much the difference between microfiche and Google in effectiveness.