1. Is your AI event logging fully automated and "always-on"?
Art. 12 mandates continuous logging during operation without manual triggers.
Yes – Logging is natively event-driven and automated
Partial – Automated in most areas, but manual gaps exist
No – Logging requires manual activation or is ad-hoc
2. Are system logs stored in a WORM (Immutable) environment?
Art. 19 requires that audit evidence be protected against tampering and unauthorized deletion.
Yes – Using S3 Object Lock, KMS Signing, or WORM hardware
Partial – IAM controls exist, but data is technically mutable
No – Standard cloud storage without integrity locks
3. Can you link an AI decision to a specific model version in < 60s?
Traceability requires model hashes and timestamps to be natively mapped in every log entry.
Yes – Fully indexed and searchable metadata
Partial – Data exists but requires manual synthesis
No – Logging schemas lack version-specific metadata