2/2 Aside from the "A large corporate will buy them out and push their own interests" concern (not possible due to the way NV Access is structured & overseen by a board who would prevent that) - what is your worry with image description and a chatbot which scours the user guide for you? (ITo be clear, such functionality wouldn't replace the existing documentation / training material etc)

Ah, gotcha! Regarding point 2, I quite literally see it as unnecessary because basic training eBooks and the full online documentation exists, so that's just a waste of resources in my opinion. My main AI/LLM concern was doing this just to attract corporate sponsorship. My issue is the sourcing, with regards to where LLM's get their data from. If this implementation will hook into an existing API, eventually, every LLM provider will require a subscription to use it, depleting your resources. Moreover, existing LLM's are simply not built for image descriptions, which is why they often get descriptions wrong or leave out information or similar. @NVAccess

The impetus for doing any of this is 100% user wants, not corporate sponsorship. Re the chat bot, yes I tend to feel the same, I'd rather refer to the manual or material - but others (probably younger than me!) like the chat bot style - so it's worth looking into. With Image recognition, like the existing add-ons, the intent would be "something is better than nothing" - not to replace people writing proper alt text - and yes, licensing of any AI model would be a key consideration

@NVAccess Not to mention instances where the expectation isn't that good alt-text should be written, but the app should be accessible from a usability standpoint, but very much isn't. But waiting on the app to become accessible isn't an option since the thing you're trying to do with *insert inacccessible app here* can't wait. But before you ask for some sighted assistance you'd like to at least get an idea of what's on the screen so you know what to ask for.