We use UKIRT and ASCA observations to determine the nature of a high-redshift (z=2.35) narrow-line AGN, previously discovered by Almaini et al. The UKIRT observations show a broad Halpha line while no Hbeta line is detected. This together with the red colour (B-K=5.4) suggests that our object is a moderately obscured QSO (A_V >3), at optical wavelengths. The ASCA data suggest a hard spectrum, probably the result of a large obscuring column, with Gamma =1.93^+0.62_-0.46, N _H ~ 10^23 cm^-2. The combined ASCA and ROSAT data again suggest a heavily obscured spectrum (N _H ~ 10^23 cm^-2 or A_V ~ 100). In this picture, the ROSAT soft X-ray emission may arise from electron scattering, in a similar fashion to local Seyfert 1.9 galaxies. Then, there is a large discrepancy between the moderate reddening witnessed in the infrared and the large X-ray absorbing column. This could be possibly explained on the basis of, e.g., high gas metallicities, or by assuming that the X-ray absorbing column is inside the dust sublimation radius. An alternative explanation can be obtained when we allow for variability between the ROSAT and ASCA observations. Then the best-fitting spectrum is still flat, Gamma =1.35^+0.16_-0.14, but with low intrinsic absorption in better agreement with the IR data, while the ROSAT normalization is a factor of 2 below the ASCA normalization. This object may be one of the bright examples of a type II QSO population at high redshift, previously undetected in optical surveys. The hard X-ray spectrum of this object suggests that such a population could make a substantial contribution to the X-ray background.