The "states shall not impose any obligations" is a great thing, but "shall ensure that ... guaranteed ..." isn't so much. Your Parliament basically tells you how to write your software, and while this particular cause may be good[1] - the general concept isn't healthy.
Granting governing bodies this level of control... that must require exceptional levels of trust in those bodies and all future ones that could be their successors. I'd argue that it's better to have the very contrary thing - a ruling that no governing body may ever dictate how one can secure their communications and how they can't (or, in more general terms - how one can write their software). Oh, and keep the "states shall not impose" clause, of course.
Yes, that leaves data-miners with their messengers still vulnerable, but I think it's less important than a general non-interference principle. And I'm for having fines for calling non-E2E messengers "secure" if the wording may confuse user into thinking it's E2E - that's basically misadvertising. The issue is information gap and "clever" marketing - fix that and things will be good.
_____
[1] It may be a problem for early prototypes, if they had started with UI/UX and a simple insecure "TCP socket server"-level stub for the messaging layer. Mandatory implementations are always a barrier.