Why inacc?

is that move worse than going e4? why? I don't see real reason, e4 makes all my pawns as a target, while c3 makes My position quite stable

Seems to me that 2. c3 is unnecessarily passive. That would appear to be the reason the engine prefers e4. I personally would probably have played c4, which also works. 2. Nc3 is also an option.

2 c3? blocks the natural square to develop Nb1.
2 e4 opens the diagonal for Bf1, prepares O-O, occupies central square e4, controls central square d5.

it's all good, but My feels about it didn't changed, I guess something is unstable about e4

2 e4 is perfectly stable: 2...Bb7 3 Bd3 f5 4 exf5 Bxg2 5 Qh5+

The computer expect e4 or d4 which are more natural in term of development. But on a human point of view, c3 is technically a reversing of the Caro Kann, so it's good, but surprising for the computer who analyze it as a passive move while you have others potential options such as c4, e4 or even Nf3

The c3 has a kibitzer average ranking of 10 out of 28 legal moves.
Which means there was at least 9 other better engine moves to pick from.
That marks the c3 move at (100% - 35.71%) = 64.29% Accurate.

A 64.29% accuracy is not generally considered an acceptable school mark, so I don't think it's acceptable for chess either, but I've played it before. If you genuinely believe that percentage is acceptable, then continue using c3.

The master database on lichess stopped at the 8 top moves and c3 was not one of them, but the lichess databaase shows the 9th being c3, matching the legal move standings.

The legal percentage is still a pass mark and the cp value is not any better or worse than the initial best move.

So I think it's premature and inaccurate to use words like inaccuracy on ply 3.

Do you like Gambits and do you think they should all be called inaccurate?

Given the opportunity to occupy a central square with a central pawn with no downside is a "where do I sign up " moment.