Catalog
[non-validated] Documented PII Sanitization in AI Training Data Processing (SF C)

[non-validated] Documented PII Sanitization in AI Training Data Processing (SF C)

Generated catalog draft. No operational assessment or human approval is claimed.

[non-validated] This question has unresolved selected work.

Question

[non-validated] Has documented PII sanitization been implemented in AI training data processing during the current review period?

Explanation

[non-validated] Has documented personally identifiable information (PII) sanitization been implemented in AI training data processing during the current 12-month review period, as required by privacy safeguards for ethical AI deployment?

Criterion and evidence

Assess documented PII sanitization in the AI training data pipeline within the current 12-month review period. Verification requires technical evidence showing implemented sanitization processes meeting AWS Prescriptive Guidance standards, including validation logs dated within the review period and engineer signoff. Absence of documentation or automated tools without human validation constitutes failure. Pending evidence must be submitted within 30 days.

AI assistance prompt

Replace this placeholder with one clear yes/no condition within the scope of AI-Oriented Ethics (Training and Implementation). Define the evidence rule and keep the question within 512 characters. Do not invent an assessment outcome.

Clarifications

What constitutes sufficient evidence for PII sanitization compliance?

Documented sanitization procedures showing implemented technical controls (e.g., hashed PII fields), validation logs dated within the review period, and engineer signoff proving human oversight. Automated tools alone without validation records are insufficient per AWS guidance.

How is missing evidence handled for this criterion?

Providers must submit technical documentation within 30 days of evidence request; pending submissions retain a 'pending' status until validated. Absence of documentation within deadline defaults to 'failed'.

References

Implementing security and responsible AI - AWS Prescriptive Guidance

https://docs.aws.amazon.com/prescriptive-guidance/latest/gen-ai-inference-architecture-and-best-practices-on-aws/implementing-security-rai.html (opens in a new tab)