I dont follow your logic.
You say they ‘need to hyper-analyse Sora output so Teams Sora cant be inside a walled garden’.
But as of today they aren’t opening Sora to Teams.
Therefore they CAN proceed without whatever data Teams users would have created.
Therefore they could just let Teams users onto Sora, and not train on their outputs, which would leave OpenAI in exactly the same position as they are now.
Seems like stupid to me