This paper focuses on the risks associated with “negative emission” techniques for drawing carbon dioxide from the atmosphere through photosynthesis and storing it in land-based sinks or underground. It examines what these risks may imply for near-term actions to limit warming to 1.5°C or 2°C above pre-industrial levels. Negative emission techniques increasingly appear – explicitly or implicitly – in discussions of options for meeting climate goals, such as those of the Paris Agreement. Negative emissions could allow society to “undo” earlier emissions, enabling us to stay within a given carbon budget in the long run, even if we exceed it early in the century. The authors identify three types of risks in counting on negative emissions: that negative emission options will not ultimately prove feasible; that their large-scale deployment involves unacceptable ecological and social impacts; and, that negative emissions activities prove less effective than hoped, either because they are subsequently reversed by human or natural forces, or because climate change impacts prove irreversible. The paper examines four main land-based negative emissions options in light of those risks: ecosystem restoration, mosaic-landscape restoration, reforestation, and bioenergy with carbon capture and sequestration (BECCS). Many mitigation pathways aiming to keep warming below 1.5°C or 2°C assume negative emissions as high as 1,000 Gt CO2. The authors find that negative emission options cannot be safely relied upon to fill such a large gap. Embarking on such pathways could lead societies to do too little to decarbonize, and greatly exceed their carbon budgets without a way to undo the damage. Pathways that rely on much smaller amounts of negative emissions, however, could prove viable, with ecosystem restoration and reforestation providing close to the required volume of negative emissions. This avoids the need to rely on other options (BECCS, in particular) that pose higher risks of technical infeasibility and unacceptable ecological and social impacts.