"It isn’t perfect, Laguna acknowledged, because the private key is stored at the server. But it’s “ten thousand times more secure” than what we use today"
This doesn't actually fix the problem.
There's two things here that need fixing - crypto between servers and client to server is something that already happens. There's nothing stopping google adopting this and we're still in exactly the same position. It's not that it isn't perfect it's that it's the same solution in different clothes.
Anyways, two things:
The end user pgp stuff needs to just be "better" - better UIs, more support, probably default support from moz, ms et al for what we consider secure today - be that mime or PGP. If the email client supports pgp natively it can guide people through creating keys when they first set up their account - the more people who use it the more secure everybody is, even if it's breakable it's still a question cost/benefit of decryting everybody's mail.
The second thing is protecting email in transit, via probably wrapping emails inside multiple layers of crypto - so your server only needs to see what domain you're sending to and only the recipient server can actually see what user it should be delivered to. This obviously has connotations for server side spam control etc but it's a price worth paying. This is easily deliverable with domains publishing pgp keys in dns or in one of the standard directories. This also neatly ties up the matter of jurisdictional control to make it only possible at the recipient side.
It is amazing that we still take risks with email that we'd never take with something like shopping on the internet. In reality we could fix it immediately by just dropping the RFCs that allow sending and receiving email in plain text and then they have to go to the provider to get access rather than just trawling the pipes which gets us some of the way there.