We present a detailed study of SMM 02399-0136, a hyperluminous, active galaxy selected from a submm survey of the distant Universe. This galaxy is the brightest source in the fields of six rich, lensing clusters, with a total area of 0.01 deg2, that we have mapped with a sensitivity of ∼ 2 mJy beam−1 at 850 μm. We identify a compact optical counterpart with an apparent magnitude of B ∼ 23 and a low surface-brightness companion ∼ 3 arcsec away. Our spectroscopy shows that both components have the same redshift; z = 2.803 ± 0.003. The emission line widths, FWHM ≃ 1000–1500 km s−1, and line ratios, along with the compact morphology and high luminosity (MB ≃ −24.0) of the galaxy, indicate that SMM 02399–0136 contains a rare dust-embedded, narrow-line or type-2 active galactic nucleus (AGN). The source is gravitationally lensed by the foreground cluster, amplifying its apparent luminosity by a factor of 2.5, and our detailed lens model allows us to accurately correct for this. Taking the amplification into account, we estimate that SMM 02399–0136 is intrinsically a factor of five times more luminous than IRAS F 10214+4724. Its far-infrared (FIR) and Ha luminosities and low surface-brightness radio emission are indicative of an interaction-induced starburst, and the star formation rate (SFR) could be several thousand solar masses per year. This assumes that the starburst is the dominant source of energy, but we cannot yet determine reliably the relative contributions of the starburst and the buried AGN. A dust mass of 5–7 × 108 M⊙ is indicated by our data for a dust temperature of 40–50 K, independent of the dominant energy source. We estimate the possible space density of such luminous submm sources, and find that while a large population of these obscured sources could be detected in future wide-field submm surveys, they are unlikely to dominate the faint counts in this waveband. Galaxies such as SMM 02399–0136 and F 10214+4724 cannot easily be detected in conventional AGN/QSO surveys, and so estimates of the prevalence of AGN in the early Universe may require significant revision.