Audit Trail

Complete history of AI interactions with classified context

6
Sessions This Week
15
Files Accessed
262
Fields Allowed
53
Fields Blocked
All sessions permanently verified on Hedera Hashgraph (enterprise ledger)
Viewing as:

Different roles see different levels of detail

Showing 6 of 6 sessions

What Gets Logged

Transparency into our audit trail design

LOGGED
User & roleJennifer Walsh, M&A
Accountability
TimestampDec 10, 10:15 AM
Timeline
Duration23 minutes
Usage patterns
Message count14 messages
Activity level
Files accessedacme_financials.xlsx
What was touched
Field counts3 allowed, 5 masked, 13 blocked
Proof controls worked
AI modelClaude Sonnet 4
Vendor tracking
Blockchain hash0x7a3f...b2c1
Immutability
NEVER STORED
Session summaryWhat you discussed
Leaks business context
Individual messagesYour questions
Privacy, liability
AI responsesWhat AI said
Privacy, liability
Actual field valuesThe $42M number
Not needed for audit
Business contextWhy you asked
Zero context retained
📋 What Auditors Actually Need

Auditors don't care about the conversation. They care about:

Who accessed what files
What was protected (field counts)
When it happened
Proof it's immutable (hash)

The actual messages? Not needed. In fact, storing them is a liability.

✨ Benefits of This Design

Privacy by Design

If someone breaches your audit logs, they get:

"Jennifer Walsh analyzed M&A documents, 13 fields blocked"

"What's the acquisition price?" → "Based on the CIM, $42M..."

Data Minimization

GDPR Art. 5 compliant — only storing what's necessary for compliance.

90% Less Storage

Full logs: ~50KB/session. Summary logs: ~2KB/session.

On-Premise AI = No Leakage

Summary generated on-prem. It describes interaction patterns, not content.

Simpler Legal Discovery

You have who, when, what was protected. No content = less liability.