7.Nbd2 is safe, but also 7.Kf1 and 7.c3 dc3 8.0-0 are to be considered. White has enough for his investment, and Black's play isn't easy at all.

I would be more concerned about the safe and sound 6...Nc5, which gives Black a good game with zero risk.

You know, if you play such gambits as white, then you do have to be reasonably speculative.

Say after 7.c3 dc3 8.0-0 cb2 9.Bxb2 Nc5, Black is two solid pawns up.

Houdini after quite some thought gives Black an advantage of one fifth of a pawn. While this can be translated to "with perfect play, Black has a small advantage", in practice it is rather "white has a dangerous initiative for his two-pawn investment (else the engine would give a much larger advantage to Black). Notice also that in such positions engines simply love material, and quite often they give erroneous evaluations.

7.Kf1 on the other hand places white's king on a clumsy square, but not for free: Black's only good reply is 7...f5, after which Black can't castle.

7.Nbd2 is absolutely safe, factly the engine gives white a very tiny advantage (one tenth of a pawn) after both 7...Ng5 and 7...Nc5. Needless to say, this safe move would NOT be my choice on the board. After all, if I wanted to play minuscule advantages, I would not play speculative gambits, on the first place!

I do play this variation (as Black), and I do play it for a win after 6...Nc5.

I would probably play 6...Bb4+ in a correspondence, engines-on game, but I doubt that I will ever get the chance. You see, since the introduction of strong engines, it's highly unlikely to meet such speculative play by your opponent over-the-board.

Well, Black can play 100% safe in this variation, which is quite easy.

...and white has absolutely nothing. If he doesn't take en-passant, he is simply worse.