Claude's reverse-engineering refusals spark debate; OpenAI says Codex blocks are accidental
sebkrier · x · 2026-10-26
A debate over safety guardrails accidentally blocking legitimate security work: users report Claude refusing to reverse engineer or remove malware from their own devices, though rephrasing gets around it — fueling Reddit rumors that Claude was nerfed. tszzl says the blocks are mostly accidental decision-boundary issues with no actual prohibition in Codex, and a fix is coming. Andrew Curran argues restricting people from working on their own property is pointless and counterproductive: users will switch to less restrictive models, costing subscriptions and goodwill — and Anthropic will likely follow OpenAI's approach.
More from Models
- Text-only LLMs can draw pixel art: Qwen3-32B best at ASCII cats and smileys — gabriberton · 2026-10-26
- The prompt behind making text-only LLMs draw 16×16 pixel images — gabriberton · 2026-10-26
- I Asked 5 Text-Only LLMs to Draw Objects They've Never Seen — gabriberton · 2026-10-26
- Japan to build AI cyberattack simulation team as AI defense goes mainstream — evijit · 2026-10-26
- Anthropic engineer: Claude Code cache hit rates are now high, savings live — teortaxesTex · 2026-10-26
- Fine-tuned 2015 CNN beats GPT-6 and Claude on skin lesion diagnosis, 62% vs 42% — lmoroney · 2026-10-26