Privacy aspects for Voice Assistants

I would not conclude the discussion the same way.
I guess most people do not know what voice processing could achieve for a company interested in exploiting it (comercially and otherwise). It also makes a huge difference if we talk about consumers or corporate users (employees). In the latter case, confidentiality is a critical requirement and off-site processing has more implications (up to the point where you need a contract).

This is an interesting interaction. I think that assistants need a comprehensive and granular consent management service. Comprehensive means it counts for all skills and features. Granular means you can control behavior, for instance:

  • per skill/feature
  • per context (e.g., location or topic)
  • globally
  • per user (assuming multi-user capabilities)
  • per device
  • based on profiles (e.g. paranoid)
  • proactive (user instructs assistent when to use Internet)
  • reactive (assistant asks if it can use Internet)