One which is absent from this discussion is the privacy declaration.
Of all the models, frameworks (MOF included) and descriptions I’ve read so far there’s no indication on what is the treatment done to user input (queries, prompts, data, etc.):
are they discarded at the end of the session?
or manually by the user?
are they safely stored?
is it available for other users?
is it available for the model?
is it used to retrain the model?
does the model/software ‘phones home’ and transmits that data?
etc, etc,…
I do think is an important aspect to be taken in account in these systems.
And although not a requirement on Open Source systems (Ai or not) it is something that should be on the recommendations for best practices.
I vaguely recall a similar topic being raised in the early discussions, and I believe I wrote something similar myself as below.
I do not think it is desirable for OSAID to address issues like privacy or ethics—matters beyond the “freedom of the system.” One major reason is that these topics are not directly related to open source. Open source has always defined freedoms that can be enjoyed equally, regardless of the country in which they are applied. In the region where I live, Japan, there are some neighboring countries with entirely different ideologies. To ensure the same freedoms can be provided even in such countries, I believe we should not incorporate concepts based on entirely different ways of thinking into the definition.