Take badly drafted and overlarded legislation and then summarise it badly, and you end up with a refined recipe for shadowbanning, blocking & censorship:

Our children’s codes for user-to-user services recommend that services … design and operate recommender systems so that content likely to be suicide or self-harm content is excluded from children’s feeds

I wonder how many errors this will trigger?

---

Via Graham:

Not convinced about the paraphrasing. The actual draft Code language feels more precautionary than ‘likely to’. #OnlineSafetyAct https://t.co/ykmDYeqhKu pic.twitter.com/d1CsXDktes

— Graham Smith ? (@Cyberleagle) August 8, 2025