moyix on LLMs' worst writing habit: "disclaiming things no one ever claimed"
moyix · x · 2026-11-20
Security researcher moyix shared a screenshot of him editing AI output, asking the model to cut phrasing that "disclaims things no one ever claimed" — with a killer example: "this should not be construed as saying we kidnapped a bunch of puppies" (nobody was wondering that, but now they are). Just be clear about what we did.
He jokes he has to find new ways to make this point daily and they're starting to sound deranged. The takeaway: defensive disclaimers in LLM text create the very suspicion they try to preempt.
More from Fun
- Gemini 4 Neon flunks Runescape Bench: ambitious strategies, terrible time management — morgymcg · 2026-11-20
- Fan makes an AI video showing how much ~30B Qwen models improved in 28 months — vanstriendaniel · 2026-11-20
- Day 4 of an AI agent run: "Whoever is in charge of mornings is doing well" — generativist · 2026-11-20
- One share filled: Anthropic IPO retail allocations are hilariously tiny — felpix_ · 2026-11-20
- GTA VI Is Out Before ICLR Reviewers Even Acknowledged Rebuttals — aran_nayebi · 2026-11-20
- Open-source orion-gaze-live predicts where you're looking from screen and input — no webcam needed — round · 2026-11-20