Event banner
Event details
Can you talk about security and compliance as it relates to features of Teams Premium that utilize large language models? Specifically, does the information that gets fed into these models (i.e., transcripts from meetings that are summarized) also serve to train these models, or is our organizational data specifically protected from that?
JohnGruszczyk
Microsoft
MicrosoftHi Dan - thanks for your question. You are correct that your organizational (customer) data remains your customer data and it's not exposed to training, customer data is not being used for modeling. Your Teams data is stored in the geographic region associated with your org/tenant and Microsoft does not have access to your meeting data. There is no change to Microsoft’s principles for commitments to data security and privacy in the enterprise. https://www.microsoft.com/en-us/ai/responsible-ai-resources
What I'd also add is that in time, ideally we would be able to have training done specific to your org/tenant (with proper data protections) in order to make AI more relevant to your org. This is not a commitment, but sharing that we do believe customers will want to see the modeling trained to make the service better specific to them/inside their environment. But again I want to be clear that customer data is not being used, I hope this was helpful
- Very helpful, thanks!