That headlines can be wrong or (intentionally) falsy stated, we have to expect. But in our private communication, it can not be accepted to get wrong summaries. Every message I receive is 100% true while reading it in raw text. I myself, by falsy reading or eventually interpreting a message, I myself make errors - but that is my incapacity. I do not need another incapacity that my phone is dyslexic about my private communication. If only this feature summarizes my message in 2-5% in a wrong summary and can not anymore always be trusted, it is something unseen before in private messaging. I have to now take the same awareness and wariness - that I have to apply to News Headlines - now also in my private messaging? I can not tolerate even 1% of wrongness, while AI seems to operate between 0-96% at best never reaching 100. Your AI should not dumb down your private communcation to a hashtag for convience of time saving but opening a can of worm of distrust for something that was previously always trustworthy.