> slippery slope arguments There is a stronger 2nd argument, but it's buried towards the end: > Now imagine an Internet of Snitches doing it badly. It’s easy

> slippery slope arguments There is a stronger 2nd argument, but it's buried towards the end: > Now imagine an Internet of Snitches doing it badly. It’s easy if you try. Some vendors already do a bad job of keeping customer data private and will continue that track record, so you can expect public leaks of databases that have flagged suspected criminals. Other vendors will write bad machine learning algorithms (or biased ones) leading to false positives and ruined lives. No slippery slope is needed, CSAM alone is extremely sensitive, a false accusation is ruinous... if automated client side CSAM scanning becomes a wide spread practice with Apple leading the way, it will be an "AI" of the flies... Apple has also demonstrated that it's willing to use ML as an unquestionable source of truth before with it's in store security, using facial recognition which is blatantly wrong to prosecute the innocent.