July 2026 Summaries

Resolução Conjunta nº 18, issued by Brazil's Conselho Monetário Nacional and Banco Central do Brasil, mandates Brazilian financial institutions to comply with 12 data-quality dimensions by December 31, 2026, making data quality a board-level obligation rather than an internal concern. This regulation aligns with BCBS 239, a global standard for effective risk data aggregation, which has shown that even well-funded global banks struggle with compliance due to legacy systems and underfunded programs. Resolução 18 requires continuous monitoring and correction of data quality issues, involving a designated director accountable to the Banco Central for overseeing processes that ensure data integrity. The regulation emphasizes the need for automation and continuous evidence of data quality across dimensions like accuracy, completeness, and traceability, turning data governance into a systematic engineering capability rather than a periodic reporting task. As the deadline approaches, Brazilian institutions are urged to implement automated checks and maintain a comprehensive audit trail to ensure readiness for regulatory scrutiny, reflecting a global trend towards stringent data governance standards.

Data contracts are essential for maintaining trust in data pipelines by ensuring that data conforms to predefined specifications, which are typically outlined in a YAML file. These contracts don't guarantee absolute correctness but ensure data consistency and reliability, thus preventing bad data from reaching end users. Soda, a data quality tool, facilitates the management and verification of these contracts through various methods, including command-line tools, the Soda Cloud platform, and agents with an MCP server. The tool allows for the creation, verification, and monitoring of data contracts while offering features like contract auto-generation and integration with AI tools for ease of use. Soda's infrastructure can be scaled according to user needs, and its commercial features include more advanced options like a Diagnostics Warehouse for tracking data issues. The central philosophy is to provide a shared interface for defining and maintaining data quality standards that all stakeholders can trust, leading to more reliable data-driven decision-making.

Data governance has evolved to meet the demands of artificial intelligence by ensuring data is "AI-ready," meaning it is trustworthy and can be acted upon by AI agents without human intervention. This shift requires governance teams to focus on data quality, ownership, and enforceability, which can be facilitated by tools like Soda AI. Soda AI helps streamline the process by automating the identification of data quality coverage, drafting data contracts, and assigning dataset ownership. These processes ensure that data meets the necessary standards for AI consumption. Despite these advancements, human oversight remains crucial as Soda AI proposes changes that require approval from governance managers, who maintain control over what is deemed "good" data. This approach helps organizations close the gap between policy and practice, particularly in regulated sectors where compliance is critical, allowing them to implement AI-driven data quality measures efficiently and effectively.

Soda offers three interfaces—CLI, API, and MCP—for executing data quality processes, all based on a single data contract written in a version-controlled YAML file. The CLI is ideal for automating repetitive checks in CI/CD pipelines, allowing data engineers to maintain data quality without extensive integration code. The API is suitable for embedding data quality checks within software applications or dashboards, providing structured JSON output for analytics engineers and internal tools. The MCP, or Model Context Protocol, leverages AI to interpret plain-language requests, making it useful for complex or large-scale tasks that require a human in the loop for approval. Each interface can be used independently or in combination, as they all share the same data contract, ensuring consistent data quality management across platforms.