If you have a self gravitating sphere than all force vectors point are normal to the surface (and other equipotentials). These force vectors only point towards the centre in the case of a sphere (or at the poles and equator of an arbitrary spheroid). However, even for a spheroid, where gravity is "assisted" by centrifugal acceleration, the gravity vector does not always point to the centre, since the centrifugal force at any point is not in general directed away from the centre, but away from the rotation axis.
The gravitational potential outside of a uniform spheroid of mass $M$ can be expressed to high accuracy with $$\Phi = -\frac{GM}{r} + \frac{kG}{2r^3}(3\cos^2\theta -1),$$ where $r$ and $\theta$ are the usual spherical coordinates and $k$ is a constant equal to the difference in the moments of inertia about axes parallel and perpendicular to the rotation axis (the object must be rotating to be a self gravitating spheroid).
Taking the gradient of this potential you see there is a term in $\hat{r}$, but also a $\hat{\theta}$ component $$g_{\theta} = \frac{3kG}{r^4}\sin \theta\cos\theta$$ that is not directed towards the centre of the spheroid.
In the limit of an almost disk-like spheroid, then at large distances from the centre and reasonably close to the disk plane, this term will dominate and the gravity will tend to act towards the disk midplane rather than towards the centre of the disk.
The above analysis only includes the monopolar and quadrupole gravity fields. There are (smaller) higher order terms that can be included, but the result is the same -- that the gravitational field outside an oblate spheroid is not a central force (Hofmeister et al. 2018).