

A former colleague at a place where I used to work added my personal MS account to a Teams community inside the organization. It split my Teams account in two, prompting me to choose which one I wanted to use every time I opened Teams.
One side was associated with the organisation, the other was still my personal account. My personal account became inaccessible and attempting to login would result in a referral loop and an error. The MS advice for the error code was to get the system admin to remove my account from the organisation, which wasn’t possible because I don’t work there anymore.
Still, people find it difficult to navigate this. Its use cases are limited, but it doesn’t enforce that limit by itself. The user needs to be knowledgeable of the limitations and care enough not to go beyond them. That’s also where the problem lies. Leaving stuff to AI, even if it compromises the results, can save SO much time that it encourages irresponsible use.
So to help remind people of the limitations of generative AI, it makes sense to fight the tendency of companies to overstate the ability of their models.