We observed VB 10 in August and September 2009 using the FORS2 camera of the VLT with the aim of measuring its astrometric motion and of probing for the presence of the announced planet VB 10b. We used the published STEPS astrometric positions of VB 10 over a timespan of 9 years, which allowed us to compare the expected motion of VB 10 due to parallax and proper motion with the observed motion and to compute precise deviations. The single-epoch precisions of our observations are about 0.1 mas, and the data showed no significant residual trend, while the presence of the planet should have induced an apparent proper motion greater than 10 mas yr-1. Subtraction of the predicted orbital motion from the observed data produces a strong trend in position residuals of VB 10. We estimated the probability that this trend is caused by random noise. After taking all the uncertainties into account and using Monte-Carlo resampling of the data, we are able to reject the existence of VB 10b with the announced mass of 6.4 MJ with a false alarm probability of only 5 × 10-4. A 3.2 MJ planet is also rejected with a false alarm probability of 0.023. Based on observations made with ESO telescopes at the La Silla Paranal Observatory under programme ID 283.C-5024.