[mtsvetanov071]

Disclaimer: Do not include SAP-internal or customer-specific information in this PR (e.g. internal system URLs, customer names, tenant IDs, or confidential configurations). This is a public repository.

Description

Several AI audit events (AiPromptReceived, AiGuardrailTriggered, AiModelInvocation, AiResponseGenerated, AiMessageSent, AiTaskCreated, AiTaskUpdated, AiUserInputRequested, AiTurnStart, AiTurnEnd) embed an AiMessageContext (proto field message_context). Two of its fields — prompt_text and response_text — hold the verbatim model input/output and are flagged in the proto as "may contain sensitive data".

This change strips those two free-text fields centrally in AuditClient.send(), before validation and serialization, so that no event type forwards the raw prompt/response content to the audit log — current or future.

Correlation fields (uid, name, ai_role_id) and the token counts (prompt_tokens, completion_tokens, total_tokens) are preserved, so events remain useful for correlation and metrics. Events without a message_context field, or with an unset one, are left untouched (the helper guards with HasField so it never auto-creates an empty sub-message).

Related Issue

Type of Change

• Bug fix (non-breaking change that fixes an issue)
• New feature (non-breaking change that adds functionality)
• Breaking change (fix or feature that would cause existing functionality to change)
• Documentation update
• Code refactoring
• Dependency update

How to Test

1. Build an AI audit event with a populated message context, e.g.:

   event = AiPromptReceived()
   event.common.tenant_id = "tenant-123"
   event.message_context.prompt_text = "sensitive prompt"
   event.message_context.response_text = "sensitive response"
   event.message_context.uid = "conversation-123"
   client.send(event)

2. Inspect the emitted record (or the in-place event after send()).
3. Expected result: prompt_text and response_text are empty; uid and token counts are retained.

Automated tests:

pytest tests/core/unit/auditlog_ng/unit/test_client.py

All 31 tests pass (5 new in TestStripMessageContextText, covering stripping, field preservation, unset contexts, events without the field, and end-to-end via send()).

Checklist

• I have read the Contributing Guidelines
• I have verified that my changes solve the issue
• I have added/updated automated tests to cover my changes
• All tests pass locally
• I have verified that my code follows the Code Guidelines
• I have updated documentation (if applicable)
• I have added type hints for all public APIs
• My code does not contain sensitive information (credentials, tokens, etc.)
• I have followed Conventional Commits for commit messages
• (New modules only) I have attached e2e evidence showing the module working against a real BTP service
• (Security-sensitive changes only) I have run /security-review and addressed all findings

Breaking Changes

None. The stripped fields are optional and were never required for a valid event. Callers that set prompt_text/response_text will simply find them absent in the emitted record; all other fields are unaffected.

Additional Notes

The scrubbing is enforced unconditionally in the single send() choke point rather than per-event, so any AI event type added later is covered automatically without further changes.

[LucasAlvesSoares]