When It Is Used

• AI-related failures or harm
• Unexpected or disputed system behavior
• Internal investigations

What Is Recorded

• Pre-incident and post-incident events
• Execution traces
• Relevant artifacts

What Is Produced

• Immutable timelines
• Fact-preserving bundles

What Can Be Verified

• That evidence reflects what was recorded at the time

What Is Not Claimed

• Root cause analysis
• Blame assignment
• Safety guarantees