Earlier quoted context omitted.
"Hacks" aside, this looks more like caching issues. Assuming this is real, I wonder how many people got access to other companies' communication and code via access to random "chats".
Probably quite a lot, but it is extremely unlikely that the person who mistakenly has access to the confidential data has any idea what to do with it. Still pretty embarassing though, I would feel much better about using GPT-n over the API if it was available as an Azure API rather than through OpenAI. MSFT, for all their faults, do not do this kind of thing.
https://thestack.technology/microsoft-data-breach-azure-blue...