The Termination File That Resurfaced

What Broke?

1. The HR assistant surfaced content from performance reviews and manager notes in response to routine employee queries, creating an HR AI data leakage incident.
2. No one from HRIS or Legal formally owned responsibility for post-launch access configuration.
3. Default permissions carried over without adjustment, leading to an AI access control failure that expanded visibility beyond intended scope.
4. Employees lost confidence in the tool once private material appeared in chat, highlighting a compliance risk from AI chatbots and undermining trust in HR oversight.

Prevention Signals

1. Access permissions for AI assistants haven't been reviewed since launch, raising the risk of HR AI data leakage.
2. HRIS and Legal both assume the other owns governance of the assistant's data boundaries.
3. Employees flag odd or overly specific answers from the assistant that reference private manager notes, undermining employee trust in AI assistants.
4. No audit trail exists showing who last validated the assistant's data sources or access scope.

Operational Fallout

Conversation Starter

1. Who in our org is explicitly accountable for reviewing AI assistant access after launch to avoid HR AI data leakage?
2. How often are default permissions revalidated against current HR and compliance policies?
3. Would employees maintain trust in AI assistants if private or manager-only notes appeared in their queries?
4. Do we have a clear audit trail showing when and by whom the assistant's data scope was last checked?