Part of me wonders whether some of the recent advancements in LLMs could help detect and stop scams by having them pre-process emails, texts, or monitor transcr

Part of me wonders whether some of the recent advancements in LLMs could help detect and stop scams by having them pre-process emails, texts, or monitor transcripts of ongoing phone calls and have them alert the user if the content seems scammy, or even just advise of common-sense precautions that could be taken (for example "you should hang up and call your bank back at the number on their website"). Another part of me is kind of horrified at the privacy implications of that, and questions if that potential trade-off would be worth it given that such a service would undoubtedly be run by a private corporation.