Which would require the somehow the audio is "intercepted" to analyze that the greeting was said. Two-party consent state include that action. Storage is another thing, no?