NBO analysis is almost meaningless for H-bond and all kinds of weak interactions, although there exists many articles tentatively using NBO to discuss H-bond.

In the NBO framework, there is no concept to reflect electrostatic interaction and dispersion interaction, which are crucial in weak interaction.

NBO only tries to reveal interaction energy corresponding to charge transfer via second-order perturbation analysis (i.e. E2 analysis), however, the result has been recently demonstrated as completely misleading, see J. Phys. Chem. A, 2017, 121 (7), pp 1531–1534.

PS: Aside from AIM, there are also many useful ways to characterize H-bond, for example, the RDG method (JACS,132,6498) and electrostatic potential analysis (J. Phys. Chem. A 2014, 118, 1697−1705), etc. Most of them can be realized by Multiwfn program (http://sobereva.com/multiwfn), see Section 4.A.5 of the manual for an overview.

I believe AIM will find a "bond" critical point between any two neighboring atoms, whether it is a bonding interaction or not.  He2 has a "bond" according to AIM.  So I would base any further discussion or analysis on the NBO results.  And this is from someone who normally prefers AIM charges to NBO charges.

NBO analysis is almost meaningless for H-bond and all kinds of weak interactions, although there exists many articles tentatively using NBO to discuss H-bond.

In the NBO framework, there is no concept to reflect electrostatic interaction and dispersion interaction, which are crucial in weak interaction.

NBO only tries to reveal interaction energy corresponding to charge transfer via second-order perturbation analysis (i.e. E2 analysis), however, the result has been recently demonstrated as completely misleading, see J. Phys. Chem. A, 2017, 121 (7), pp 1531–1534.

PS: Aside from AIM, there are also many useful ways to characterize H-bond, for example, the RDG method (JACS,132,6498) and electrostatic potential analysis (J. Phys. Chem. A 2014, 118, 1697−1705), etc. Most of them can be realized by Multiwfn program (http://sobereva.com/multiwfn), see Section 4.A.5 of the manual for an overview.

AIM shows bond paths even when atoms are not bonded. So AIM  is seductive but also somewhat misleading. NBO is an EDA method provides an energy perspective of the interaction. It at least tells you if a given interaction is stabilizing or not. There are other EDA methods (e.g., BLW), which give you the energy components of a given interaction. Generally using both EDA and AIM provides complementary information.

Sason

It is wrong to draw chemical conclusions merely from the existence of a "bond" path / "bond" critical point (bcp) between two atoms in AIM. You should look at the values of the electron density and Laplacian at the bcp (and possibly localization/delocalization indices) before drawing any conclusions from AIM.

Regarding the comments that the QTAIM sees "bonds" even when there is "no bond" check my paper: Article Why Bond Critical Points Are Not "Bond" Critical Points

Comparing AIM to NBO is like comparing apples to models of oranges. Instead of churning-out meaningless answers, try to ask better questions.