ISO 42001 A.11.4: Protecting Personal Data in AI Systems
As AI systems increasingly process personal data, ISO 42001 A.11.4 requires organizations to implement robust controls ensuring lawful, ethical data handling. This control bridges AI governance and data protection compliance—critical for SMBs deploying AI while respecting privacy regulations like GDPR.
What this means
Control A.11.4 mandates that organizations establish and maintain controls protecting personal data throughout AI system lifecycles. This includes securing a lawful basis for processing, minimizing data used in AI training to only what's necessary, enabling data subject rights (access, deletion, portability), and aligning AI data practices with applicable privacy laws. The control recognizes that AI amplifies data risks—models can memorize sensitive information, inference may violate privacy, and outputs can expose personal details—requiring preventative governance.
How to comply
- 1.Document lawful basis for all personal data processing in AI systems (consent, contractual necessity, legal obligation, vital interests, public task, or legitimate interests)
- 2.Conduct data minimization reviews before AI training—collect and retain only data necessary for model objectives
- 3.Implement technical controls to prevent AI systems from inferring or exposing protected personal information
- 4.Establish processes for data subject requests (access, correction, deletion, portability) and execute within legal timeframes
- 5.Map all AI systems processing personal data and classify by data sensitivity and processing purpose
- 6.Maintain data retention schedules aligned with AI model refresh cycles and legal requirements
- 7.Train AI teams on data protection principles and privacy-by-design practices
- 8.Conduct data protection impact assessments (DPIAs) for high-risk AI systems before deployment
Evidence auditors look for
- Data inventory documenting lawful basis for each AI system's personal data inputs
- Data minimization audit report showing reduction of unnecessary features in training datasets
- Privacy policy addendum explaining how personal data is used in AI systems and subject rights
- Technical documentation of de-identification or anonymization techniques applied pre-training
- Data subject request procedure and fulfillment logs (access, deletion, portability requests)
- Data protection impact assessments (DPIA) for AI systems processing sensitive personal data
- Training records showing staff awareness of personal data protection in AI contexts
- Data retention and deletion logs for personal data used in model development and inference
Frequently asked questions
When will FAQs be available?
The FAQ for this control is currently being prepared.
See how GRCWatch handles this control automatically.
See how GRCWatch handles this control automatically
Start free trial