Training opt-outs may be a loophole: opted-out data could still improve models
logicus · x · 2026-09-13
- logicus argues that "training opt-outs" in terms/privacy policies are ambiguously worded and could imply various derivative uses of user data that technically honor the no-training promise.
- Opted-out data could effectively be transformed into training signals that teach models the skills and knowledge a user displays in interactions.
- Example: authors or thinkers using AI may progressively train the model to write and think like them. "It just seems dark."
- Thread context: debate over whether any vendor offers a truly ironclad opt-out.
Related event: Developers Question Whether AI Training Opt-Out Clauses Leave Loopholes(3 posts)→
More from Models
- Dev reality check: hours babysitting frontier model errors vs X hype of superhuman AI — mervenoyann · 2026-09-13
- DeepSeek-V4.1-Flash reportedly live on Ollama cloud with zero data retention and off-peak pricing — johnseach · 2026-09-13
- Running two frontier models against each other on study design works 'absurdly' well — Tkaraletsos · 2026-09-13
- DeepSeek 4.1 Flash solves HLE problem, then downloads the dataset to check its own answer — FutureStriking283 · 2026-09-13
- Anthropic's policy details exactly when your Claude chats are used for model training — austinc3301 · 2026-09-13
- GPT-Live-1 first impressions: most natural voice model yet, but instruction following is unreliable — kolchinski · 2026-09-13