Aaronson specifically addresses this in section 1.2.1, on page 6.
Internet lore and popular media then assume that immediately for example all encryption will be trivially broken. Mathematically it may be true because then all NP problems would be known to be polynomial, but there's still the issue of the practical steps involved factorising a huge number which may have enormous constants or very large exponents. Or if it turns out we can reduce it to some other known polynomial problem, there's still the actual transformation which itself may be polynomial with large constants or exponents.
Also, as others have pointed out, the constants involved in any "generic" algorithm for solving any NP problem in Polynomial type may be astronomical, so even if it turns out that P=NP, then it may not ever be feasible to actually such any algorithm for anything practical.
Maybe you meant that factorization hasn't been proven to be NP-complete, which is true. In fact it would be very surprising if it was NP-complete, because that would imply other surprising things (e.g. NP = co-NP).