Access controls for copied or derived assets are not fixed or uniform; they follow a flexible propagation approach. When an asset is copied, the new asset can start by inheriting the existing permissions from the source, providing continuity of access. At the same time, those permissions may be re-evaluated to fit the new context, especially if the derived asset has a different risk level or exposure. In some cases, the derived asset retains the parent permissions; in others, governance workflows are triggered to review and adjust access before granting or denying permissions. This combination—inherited permissions, possible re-assessment, and governance-driven adjustments—captures the real-world behavior of access control propagation. Access rules aren’t reset automatically in every case, and access is indeed propagated rather than never carried over. Also, inheritance typically involves both users and groups, not just users, so limiting it to users isn’t correct.

Foundry Aware treats data assets as the managed building blocks you work with in the platform: datasets and tables that hold data, notebooks that contain analysis code, pipelines that describe data transformations, apps that expose functionality, and catalog entries that register and describe these assets. Dashboards, reports, and data visualizations are outputs built on top of these assets rather than the assets themselves. External memory caches and temporary tables are ephemeral runtime constructs not stored as persistent assets in the catalog. Physical files without metadata lack the governance and discoverability that define a data asset. So the data assets you work with are datasets, tables, notebooks, pipelines, apps, and catalog entries.

Cross-region governance in Foundry Aware is built to respect local rules while enabling safe collaboration across regions. Region-Specific Policies are the backbone here, enforcing access controls, data processing rules, retention, and compliance requirements that vary by region. This ensures that what you can do with data in one region aligns with that region’s laws and agreements. Replicated Catalogs support governance by keeping metadata synchronized regionally. Teams can discover and catalog assets locally, maintaining visibility and control without pulling data across borders unnecessarily. This balances local autonomy with a coherent governance framework. Compliant Data Transfer provides the secure, auditable pathways needed to move data between regions when allowed. It covers encryption, provenance, checks, and logging, ensuring transfers meet regulatory and policy requirements. Other options miss essential parts: a single global policy can’t accommodate regional legal differences, having no governance leaves data unprotected and noncompliant, and making governance optional undercuts accountability and control.

Data contracts define what data exchanges must provide in terms of structure, meaning, and delivery guarantees. In practice, that means three pieces: schemas, which describe the data's shape—field names, types, and nesting; semantics, which specify what each field represents, allowable values, units, and any business rules or validations that ensure the data is interpreted correctly; and SLAs, which lay out the operational expectations for the data flow—how often data is produced, how fresh it needs to be, and the required reliability and latency. The combination ensures downstream systems know exactly what to expect and how to rely on it. If you only specify the structure, you might miss the business interpretation and the timing guarantees; if you only specify semantics, you may know what the data means but not how it's organized or delivered; and without SLAs, there are no explicit commitments about delivery cadence or uptime. For example, a customer feed might require a schema with fields like customer_id, balance, and last_updated; semantics that customer_id is a unique identifier, balance is a non-negative number in USD with two decimals, and last_updated is the timestamp of the record; and SLAs stating updates every 15 minutes with latency under 2 minutes and 99.9% uptime.