Locket – Robust feature-level access control for LLMs

Locket introduces a security framework for LLMs that enforces granular, feature-level access control—such as restricting coding or customer support capabilities—enabling use cases like A/B testing, content/age restrictions, and pay-to-unlock monetization. It is designed for developers and platform operators who need to safely segment model functionality without deploying separate models. This is interesting because it bridges the gap between monolithic LLM deployments and the fine-grained, policy-driven access required for production security and business logic.