Recently, gas disks have been discovered around main sequence stars well beyond the usual protoplanetary disk lifetimes (i.e., 10 Myrs), when planets have already formed 1-4 . These gas disks, mainly composed of CO, carbon, and oxygen 5-7 seem to be ubiquitous 3 in systems with planetesimal belts (similar to our Kuiper belt), and can last for hundreds of millions of years 8 . Planets orbiting in these gas disks will accrete 9, 10 a large quantity of gas that will transform their primordial atmospheres into new secondary atmospheres with compositions similar to that of the parent gas disk. Here, we quantify how large a secondary atmosphere can be created for a variety of observed gas disks and for a wide range of planet types.We find that gas accretion in this late phase is very significant and an Earth's atmospheric mass of gas is readily accreted on terrestrial planets in very tenuous gas disks. In slightly more massive disks, we show that massive CO atmospheres can be accreted, forming planets with up to sub-Neptune-like pressures. Our new results demonstrate that new secondary atmospheres with high metallicities and high C/O ratios will be created in these late gas disks, resetting their primordial compositions inherited from the protoplanetary disk phase, and providing a new birth to planets that lost their atmosphere to photoevaporation or giant 1 arXiv:2004.02496v1 [astro-ph.EP] 6 Apr 2020 impacts. We therefore propose a new paradigm for the formation of atmospheres on lowmass planets, which can be tested with future observations (JWST, ELT, ARIEL). We also show that this late accretion would show a very clear signature in Sub-Neptunes or cold exo-Jupiters. Finally, we find that accretion creates cavities in late gas disks, which could be used as a new planet detection method, for low mass planets a few au to a few tens of au from their host stars.The discovery of large amounts of gas around main-sequence stars is recent with most detections occurring in the last few years 3, 11 . These late gas disks are observed in systems that have planetesimal belts, which are older than 10 Myr and can last for hundreds of millions of years 8 .It is thought that the observed gas is released from volatile-rich planetesimals when they collide with each other in the system's belts 6, 12 . The gas then viscously evolves 4 , spreading inward and outwards 10, 15 . Hence the observed gas is likely secondary (rather than of primordial origin) and this late disk (main-sequence) phase is different from the younger (<10 Myr) protoplanetary disks that are much more massive, hydrogen-rich and in which giant planets form within a few millions of years 16 .These late gas disks are nearly ubiquitous around A-type stars; Gas has been detected around more than 70% of systems with bright planetesimal belts 3 . As for other stellar type stars or lower mass systems the statistics are still based on too small a sample as these gas disks are harder to detect but gas evolution models 6 predict that all stars surrounded by planetesimal belts s...