Asking Opus "what do you think?" gets flagged as a distillation attack
Dany0 · reddit · 2026-09-29
Reddit user Dany0 complains about Anthropic's Opus over-refusing: simply asking the model "what are your thoughts on this?" was blocked by safety guardrails as an alleged "distillation attack" (screenshot included).
Written in tongue-in-cheek LocalLLaMA style, the post jokes about where the "Opus 5.5 datasets" are and mocks how frontier models' guardrails now flag perfectly normal prompts.
More from Fun
- David Patterson mocks AI safety rhetoric: encyclopedias contain bomb plans, so they're 'misaligned' — davidpattersonx · 2026-09-29
- No Kimi launch this week, says leaker ChrisGPT, cooling API rumors — ChrisGPT · 2026-09-29
- AI Twitter goes gym-brained: "re-rack your weights" meme — AIFlow_ML · 2026-09-29
- Instinct CEO's agent pitch: restaurants prioritizing birthdays via agents gets mocked — himanshustwts · 2026-09-29
- Dev rants that AI news media are 'controlled ops' and influencers are 'low-IQ shills', vows to fix it — ns123abc · 2026-09-29
- 77-year-old grandpa builds vertical farming business plan with ChatGPT on a flight, calls AI 'sir' — theteknosaur · 2026-09-29