This paper presents a control strategy that brings the monitoring and enforcement of flight-envelope constraints in launch vehicles down to the GNC level. This technology, widely known as safety filters in robotics and the automotive sector, serves two main purposes: to provide a formal methodology of addressing safety-critical state constraints, and to remain modular and minimally invasive to existing GNC architectures. Yet, many additional advantages stem from the integration of safety filters in launchers. They offer a substantial reduction in validation and verification, and missionization efforts; they ensure increased launch opportunities under stronger weather conditions, and they allow a redistribution of efforts between computational guidance and control. The safety filter proposed in this paper is based on a robust version of high-order control barrier functions, which are formulated as efficient, quadratic optimization problems that account for unknown, but bounded, wind disturbances. Moreover, it produces residual signals that carry essential information for safety monitoring, autonomous flight termination capabilities and guidance re-computation algorithms. We also present a relevant tradeoff between stability and safety manifested in launchers and other unstable systems, which highly influences the design of the safety filter. Finally, we demonstrate the effectiveness of this technology in a Monte Carlo campaign using a high-fidelity simulator.