Probably for the same reason as slide 2 (Safe Harbour: this is me, not Oracle, not a commitment, etc) in http://llvm.org/devmtg/2014-10/Slides/Hawes-Frappe.pdf from Oracle: fear that anything that might get said will lead to legal proceedings. The slides will have been vetted, but you cannot prevent a speaker from ever saying 'we do [not] plan' (might get interpreted as "we, Apple") or something like it.
I want to listen to this talk so much, maybe the will go over it and release it later. Until now JIT and LLVM were not the best match. But the safepoint work will really make things easier for managed languages and things like FTL will probebly provide good examples of how to use LLVM in that context.