LLMs Are the Key to Mutation Testing and Better Compliance

Meta's Automated Compliance Hardening (ACH) tool leverages large language models to overcome traditional barriers in mutation testing, enabling scalable deployment at industrial scale. The system generates realistic, domain-specific mutants and corresponding tests through plain-text prompts, addressing longstanding challenges like computational expense, equivalent mutants, and unrealistic fault simulation. During a three-month trial across Meta's platforms, engineers accepted 73% of generated tests, with 36% deemed privacy-relevant. The approach transforms compliance from manual, error-prone processes into automated systems that catch bugs before production while reducing cognitive load on developers.