Power BI AI Audit Logging

When AI changes a Power BI model,
you need to know exactly what happened

SemanticOps records MCP activity, model operations, policy decisions, and outcomes so teams can review what happened during AI-assisted development.

Important scope note

SemanticOps audit logging covers SemanticOps MCP activity — the tool calls, model operations, policy decisions, and test results that flow through the SemanticOps server. It does not replace Microsoft tenant-level audit logs and does not capture edits made directly in Power BI Desktop, Tabular Editor, or other tools that do not use the SemanticOps MCP. That scope boundary is intentional and stated clearly in the docs.

Log contents

Eight categories recorded in every log entry

Each log entry captures who initiated the action, what model was involved, what operation ran, what object changed, whether the action succeeded, what the policy decided, and whether tests ran.

Actor

User, client, assistant context

Model

Workspace, dataset, local file

Operation

Read, query, update, delete, test, rollback

Object

Measure, column, table, relationship, role

Outcome

Success, failure, blocked

Policy decision

Allowed, denied, required confirmation

Test result

Passed, failed, skipped

Integrity

Tamper-evident verification

Example entries

What audit entries look like

2026-05-23 10:42
ClientClaude Code
ModelSales Analytics
ActionDelete measure [Gross Margin %]
PolicyBLOCKED — downstream dependencies detected
OutcomeNo change applied
2026-05-23 11:05
ClientVS Code
ModelCustomer Model
ActionExecute DAX query
PolicyALLOWED WITH MASKING
MaskedCustomer Email, Customer Name
OutcomeMasked before assistant context
2026-05-23 11:21
ClientCursor
ModelFinance Analytics
ActionModify RLS role [Sales Manager]
PolicyREQUIRE TEST
Test suiteRLS regression pack
TestsFailed 2 / 18 tests
OutcomeDeployment blocked

Know exactly what AI did to your semantic model.

Tamper-evident audit logs for every MCP tool call, policy decision, and test result.