News for hackers. Stuff that matters.
Re:Anthropic reverses privacy stance, will train on Claude c... by I_am_tiberius on Friday August 29, 2025 @08:03AM (#45062936)
"five year retention". If it's in a model once, it's there forever.
Re:Anthropic reverses privacy stance, will train on Claude c... by Hnrobert42 on Friday August 29, 2025 @08:12AM (#45063032)
Is that true? Do models get rebuilt from scratch each time or do they get iterated on?
Re:Anthropic reverses privacy stance, will train on Claude c... by I_am_tiberius on Friday August 29, 2025 @08:15AM (#45063058)
I believe the big models currently get built from scratch (with random starting weights). That wasn't my point though. I meant a model created once, might be used for a very long time. Maybe they even release the weights at one point ("open source").
Re:Anthropic reverses privacy stance, will train on Claude c... by whimsicalism on Friday August 29, 2025 @09:41AM (#45064024)
yes, it’s a very big loophole. and if it’s a generative model, you can just launder the data through synthetic generation/distillation to future models
Re:Anthropic reverses privacy stance, will train on Claude c... by disconcision on Friday August 29, 2025 @10:49AM (#45064875)
this is somewhat true but i'm not sure how load bearing it is. for one, i think it's going to be a while until 'we asked the model what bob said' is as admissible as the result of a database query