Issue description:Detailed Description: We are encountering an issue in the Dynamics 365 Contact Center where the AI transcription flags foul language that the agent did not speak, and which is also not present in the call audio recording. Environment: Production Date of First Occurrence: After Go-Live on 29th August Issue Details: The transcription incorrectly inserted inappropriate/foul language in the conversation text. Review of the actual call audio confirms the agent never used such language, and no such words exist in the recording. This indicates the transcription service is generating false positives. The issue is intermittent and based on user reports; attached screenshot illustrates one occurrence. Full repro steps are not available, as the issue does not happen consistently.

 Conversation ID https://mywestminster.crm4.dynamics.com/main.aspx?appid=54ddfa90-b011-eb11-a812-000d3ab2e7b9&pagetype=entityrecord&etn=msdyn_ocliveworkitem&id=E5045BC5-4F3E-41C5-860B-EFD80AAEB473&formid=136b3800-234b-420c-9d8a-458b61cb2263 Impact: Produces misleading and inaccurate conversation transcripts. Raises unnecessary compliance concerns and affects the credibility of reporting and agent performance monitoring. Creates audit risks if transcripts are used for compliance/legal reviews. Request: Please investigate why transcription is detecting foul language when it is not spoken and not present in the audio. Provide a fix or mitigation guidance to prevent recurrence.