ISO 42001 A.7.2: Data Quality Requirements for AI Systems
AI systems are only as reliable as the data they're trained on. Control A.7.2 requires you to define and maintain rigorous data quality standards—including accuracy, completeness, timeliness, consistency, and representativeness—to ensure your AI delivers safe, predictable results aligned with your organization's actual needs.
What this means
This control mandates that your organization establish clear data quality criteria before deploying any AI system. You must actively measure and maintain these standards across all datasets used in AI applications. Quality dimensions encompass accuracy (how correct the data is), completeness (whether required fields are populated), timeliness (how current the data remains), consistency (uniform formatting and logic), and representativeness (whether the data reflects the actual population or use case the AI will encounter). The control emphasizes that quality standards must be tailored to each AI system's intended use—generic one-size-fits-all approaches fall short.
How to comply
- 1.Document data quality requirements for each AI system, specifying acceptable thresholds for accuracy, completeness, timeliness, consistency, and representativeness
- 2.Establish baseline measurements of your current datasets against these quality criteria
- 3.Implement data profiling and validation tools to continuously monitor quality across all AI training and operational datasets
- 4.Define data ownership and assign responsibility for maintaining quality standards throughout the AI system's lifecycle
- 5.Create procedures to identify, log, and remediate data quality issues before they degrade AI performance
- 6.Conduct regular quality audits and document results as evidence of ongoing compliance
- 7.Update data quality requirements whenever AI system use cases or operational contexts change
Evidence auditors look for
- Data quality policy document defining accuracy, completeness, timeliness, consistency, and representativeness criteria by AI system
- Data quality baseline assessment reports showing measured quality metrics before AI deployment
- Data validation and profiling logs demonstrating continuous monitoring
- Data quality incident reports and remediation records
- Training dataset documentation confirming representativeness for intended use cases
- Quality metrics dashboards or reports reviewed at governance meetings
- Data stewardship assignments and ownership documentation
- AI system update logs triggered by data quality concerns
Frequently asked questions
When will FAQs be available?
The FAQ for this control is currently being prepared.
GRCWatch automates data quality monitoring across your AI datasets by integrating data profiling rules, tracking quality metrics in real time, and alerting you to representativeness and completeness gaps before they degrade model performance—eliminating manual audits and keeping A.7.2 compliance continuously enforced.
See how GRCWatch handles this control automatically
Start free trial