This is merely a preprint, not a published paper. In other words, this has not been referred – subjected to regular scientific scrutiny.
Preprints are of great interest to researchers in the field -- they give them quick access to recent results before the slower process of scientific verification takes place. But preprints are not published papers (even those are not all correct!) – they aren't really useful to the general public. Especially in the case of major open problems like P=NP, such extraordinary claims require extraordinary verification, and this has yet to take place.
The submission headline here is very misleading, as is the summary. Either this preprint is correct (extremely unlikely), and then it definitely shows that P!=NP, or the preprint if wrong (almost surely the case), at which point it's not clear that it contains enough correct and deep results to actually suggest anything about whether P=NP or not. A much better headline would have been "new arXiv preprint purports to prove that P!=NP", and a better summary would have been
A recent arXiv preprint by Norbert Blum of the University of Bonn claims to show that P!=NP. This work has not been vetted by the general community and (as with every other claim of this type) is generally assumed to be incorrect. Readers who are not experts in complexity theory are advised to ignore this preprint until experts have had time to examine this work and its implications.