Active galactic nuclei (AGNs) and quasars often exhibit infrared excesses at lambda = 2-10 mu m attributable to thermal dust emission. In this paper we propose that this hot dust is supplied by circumstellar mass loss from evolved stars in the nuclear star cluster. The physics of the mass-loss dust, specifically the evaporation temperature, is a critical parameter in determining the accretion rate of mass-loss material onto the central AGN. For standard interstellar dust grains with an evaporation temperature of 1800 K the dust is destroyed inside a radius of 1 pc from a central luminosity source of 5 x 10 L.. The mass-loss material inside 1 pc will therefore have a lower radiation pressure efficiency and accrete inward. Outside this critical radius, dust may survive, and the mass loss is accelerated outward owing to the high radiation pressure efficiency of the dust mixed with the gas. The outflowing material will consist of discrete trails of debris shed by the individual mass-loss stars, and we suggest that these trails produce the broad absorption lines (BALs) seen in 5%-10% of QSOs. The model accounts naturally for the maximum outflow velocities seen in the BALs (similar to 30,000 km s(-1) and varying as L(1/4)) since this maximum terminal velocity occurs for matter originating at the inner edge of the radiative equilibrium dust survival zone. Although the radiation pressure acts on the dust, individual grains will be highly charged (Z similar to 10(3)+), and the grains are therefore strongly coupled to the gas through the ambient magnetic fields. Numerical hydrodynamic calculations were done to follow the evolution of mass-loss material. As the orbiting debris is driven outward by radiation pressure, the trail forms a spiral with initially high pitch angle (similar to 85 degrees). The trails are compressed into thin ribbons in the radial direction-initially by the radiation pressure gradients due to absorption within the trail. After reaching >10(4) km s(-1) radial velocity, the compression can be maintained by ram pressure due to an ambient gas of modest density (similar to 10(2) cm(-3)). Each of the stellar contrails will have mean column density similar to 10(19)-10(21) cm(-2), volume density similar to 10(8)-10(9) cm(-3), and thickness 10(11)-10(12) cm along the line of sight to the AGN-corresponding to parameters deduced from observations of the BAL clouds. Assuming minimal expansion perpendicular to the line of sight at the speed of sound, the width of the trails is 10(15)-10(16) Cm, Or 10(2)-10(3) times the line-of-sight depth. Since the UV-emitting accretion disk probably has a radius of about 2 x 10(16) cm, a single trail will only partially cover the continuum, but for the column densities quoted above the observed absorption lines (e.g., C IV) will be optically thick with 1.>10. Since the contrails are nearly radial just after leaving the star when the maximum outward acceleration occurs, a large range of velocities (similar to 4000 km s(-1)) will be seen in absorption of the QSO light from each trail, and only a few disk-crossing trails are needed to account for the full width of broad absorption line troughs.