Cryptographer Matthew Green: Bypassing model guardrails is trivially cheap
matthew_d_green · x · 2026-11-09
Cryptographer Matthew Green argued on X that abliteration—the technique of stripping safety guardrails from models—is cheap to perform, comparing it to the difference between driving an automatic versus a manual car, implying that defeating model safety layers poses little barrier. This came in response to a researcher noting that developing zero-day exploits with GLM-5.3 or DeepSeek-V4.1-Flash-1017 doesn't even require abliteration. The exchange reflects widespread skepticism about the practical effectiveness of open-model safety guardrails.
More from Models
- SimpleHC cuts Hyper-Connections training overhead from 59–100% to just 2.4% — SonglinYang4 · 2026-11-09
- Unverified claim: Opus 5.5 has only 10B active params; dev hands projects to Qwen-4-Flash — teortaxesTex · 2026-11-09
- Benchmarks are like perfume: why public leaderboards demand skepticism and private evals — intellectronica · 2026-11-09
- Running 8-25 Opus 5.5 threads all day used under 20% of a $200 Claude sub — cyrus_zei · 2026-11-09
- Grok Bot quietly throttled usage, then reset quotas after user backlash — xiaohu · 2026-11-09
- Two padding tokens doubled a hosted DeepSeek V4.1-Flash success rate, pointing to phase sensitivity — No_Pea_6625 · 2026-11-09