A Policy Relationship Model for Consistent Data Product Governance

The proper use of a data product requires the definition of contracts that ensure alignment with the data governance policies adopted by the organization that is offering such data product. These contracts are typically defined using natural language and later enforced through the Policy-as-Code paradigm. However, such policies do not exist in isolation, as they may exhibit relationships with one another that must be identified and resolved to ensure consistency and enable correct enforcement. Currently, this process largely relies on manual analysis, making it time-consuming and error-prone. This work addresses this limitation by introducing a Policy Relationship Model, which first formalizes the notion of data product policies and then classifies the possible types of relationships among them within the data sharing context. Building on this model, we propose a module for self-serve data platforms that leverages this knowledge to support data owners by automating key steps in the policy lifecycle. In particular, the module first translates policies into a structured, executable representation in Rego, which is then used to identify relationships between new and existing policies. To further reduce manual effort, we investigate the use of Large Language Models (LLMs) as automation tools. An initial implementation of the proposed approach is evaluated, with a particular focus on the generation of executable policies from natural language specifications.