Microsoft's new safety system can catch hallucinations in its customers' AI appsAzure AI Studio tools can detect malicious prompt attacks and unsupported responses.New safety features such as Prompt Shields and Groundedness Detection aim to enhance security for Azure AI customers.