What we've heard from customers

What we’ve heard so far:

• White-listing/black-listing by users, groups, LDAP groups, teams, etc

• Taylor (AMPAC) - a list of users (e.g. whitelist) or a AD/LDAP group - like how I can assign AD groups to channels

  • Current setup for team restrictions will work for now - can workaround with a ‘fake team’

• Britanny/Rohit (ServiceNow) - They will likely need user access controls in order to roll out in production

  • Rohit is unclear on exactly what limitations will be mandated by their risk assessment team until the review is complete. It will likely be based on users or user attributes - for example, Rohit mentioned they may need to ensure HR folks can’t access AI features. They use LDAP as their IDP

• Kakaobank - love the multi-LLM concept, but need support to hide their local LLM when MM is accessed from outside IP addresses. (Notes)

  • if there is no such way to control the AI plugin, there is no chance to sell AI plugin in public and financial sector because using personal devices to connect the MM server is common and by law, AI bot is not allowed to be appeared in public IP

• Campbell - interested in AI search, but wary of LLMs exposing data improperly. They need user access controls to apply to AI-generated search. “Ensure search returns only what you have access to”.