Report to Ombuds is processed with “AI” without consent

[hidde]

I think this may not have been the case before when I tried earlier, but today I saw this interaction on the Ombuds platform… when a user submits their case on w3c.tequitable.com, they are shown immediate recommendation actions, which when I submitted something as a test looked liked this:

In the right bottom corner it says “powered by AI”. It is unclear what the “AI” is (I'm assuming an LLM?), so I opened this issue for us to find out:

• is the user input sent to an external LLM vendor, if so, which one?
• which LLM is used and what kinds of properties of it are available? Eg, how much bias mitigation was done?
• if it would be possible for us to customise functionality?

I wonder if:

• we can add more information about the specific LLM folks are interacting with
• we should consider having this functionality be turned off for W3C, as the responses seem generic and potentially AI generated

[TzviyaSiegman]

Thanks @hidde. I've asked Catrina to look into this as our liaison with tEQuitable.