Keep in mind that this depends entirely on the map projection that you’re using. The picture you’re showing appears to most likely be a Mercator projection, which is notorious for that type of distortion, especially near the poles.
Basically all aviation charts use a Lambert conformal conic projection. While not a definitive source, Wikipedia gives the short version of this (Lambert conformal conic projection - Wikipedia):
Pilots use aeronautical charts based on LCC because a straight line drawn on a Lambert conformal conic projection approximates a great-circle route between endpoints for typical flight distances.
All ways of attempting to “flatten” a spherical (or ellipsoid) surface onto a flat map are not perfect and have various failings like distortions of some form or fashion. For example, the sample picture on Wikipedia for the LCC looks like this (where you can see that the distortion gets fairly extreme towards the southerly parts of it):
In your New York to Moscow example, as the route of flight has a significant amount of it north of the 50N standard parallel, using this particular map would deviate quite a bit from the great circle route, and one that was made using standard parallels farther north for the projection would be noticeably more accurate. However, both would be a lot closer than if you used the Mercator projection. (Note that most of your normal aviation charts are essentially small bits of the sample picture with the standard parallels chosen to minimize distortion for that particular region, so the distortion isn’t as severe as if you were trying to go from one end of the sample picture to the other.)