We applied the vertical Jeans equation to the Milky Way disk in order to study non-axisymmetric variations in the thin disk surface density. We divided the disk plane into area cells with a 100 pc grid spacing and used four separate subsets of the Gaia DR3 stars, defined by cuts in absolute magnitude, that reach distances up to 3 kpc. The vertical Jeans equation is informed by the stellar number density field and the vertical velocity field; for the former, we used maps produced via Gaussian process regression; for the latter, we used Bayesian neural network radial velocity predictions, which allowed us to utilise the full power of the Gaia DR3 proper motion sample. For the first time, we find evidence of a spiral arm in the form of an over-density in the dynamically measured disk surface density, detected in all four data samples, which agrees very well with the spiral arm as traced by stellar age and chemistry. We fitted a simple spiral arm model to this feature and infer a relative over-density of roughly 20% and a width of roughly 400 pc. We also infer a thin disk surface density scale length of 3.3–4.2 kpc when restricting the analysis to stars within a distance of 2 kpc.