I think we'll see research into how to appraise AI models for what has been weighted. Whether a model creates objectively fair reasonings or creates something harmful. Also data democratization of these models (having them available outside of large institutions) will be important. At the same time harmful: A comment a few down is mentioning how these things will be used to create AI influencers, or how propaganda campaigns could be entirely automated and deployed against other countries. Very real threats. :(