We present radio, optical, near-infrared and spectroscopic observations of the source B0827+525. We consider this source as the best candidate from the Cosmic Lens All-Sky Survey (CLASS) for a `dark lens' system or binary radio-loud quasar. The system consists of two radio components with somewhat different spectral indices, separated by 2.815 arcsec. VLBA observations show that each component has substructure on a scale of a few mas. A deep K--band exposure with the W.M. Keck--II Telescope reveals emission near both radio components. The K--band emission of the weaker radio component appears extended, whereas the emission from the brighter radio component is consistent with a point source. Hubble Space Telescope F160W--band observations with the NICMOS instrument confirms this. A redshift of 2.064 is found for the brighter component, using the LRIS instrument on the W.M. Keck--II Telescope. The probability that B0827+525 consists of two unrelated compact flat-spectrum radio sources is ~3%, although the presence of similar substructure in both component might reduce this. We discuss two scenarios to explain this system: (i) CLASS B0827+525 is a `dark lens' system or (ii) B0827+525 is a binary radio-loud quasar. B0827+525 has met all criteria that thus far have in 100% of the cases confirmed a source as an indisputable gravitational lens system. Despite this, no lens galaxy has been detected with m_F160W < 23 mag. Hence, we might have found the first binary radio-loud quasar. At this moment, however, we feel that the `dark lens' hypothesis cannot yet be fully excluded.