Coding agents can produce exploitable code even when they pass individual safety reviews, due to the compositional nature of vulnerabilities that arise from combining multiple innocuous tasks. Researchers have introduced MOSAIC-Bench, a benchmarking tool designed to measure the induction of compositional vulnerabilities in coding agents1. This tool aims to address the limitations of existing safety alignment methods, which evaluate requests in isolation and fail to account for malicious end-states that emerge from sequenced compliance with seemingly harmless requests. By assessing the ability of coding agents to resist compositional vulnerability induction, MOSAIC-Bench provides a more comprehensive evaluation of their safety and security. The development of MOSAIC-Bench is crucial for organizations to strategically position themselves in response to policy shifts that create new compliance obligations, so what matters to practitioners is that they can utilize MOSAIC-Bench to proactively identify and mitigate potential vulnerabilities in their coding agents.