Crypto professor spars with security researcher: is abliterated GLM actually weaponizable by non-experts?
matthew_d_green · x · 2026-11-09
Johns Hopkins cryptographer Matthew Green clapped back at security researcher MalwareTechBlog over the dangers of abliterated GLM (a guardrail-stripped open model). Green accused him of confidently telling hundreds of thousands of followers that no non-expert could develop a useful exploit with it — while never having directly worked with the model — and told him to either stop or go verify it himself. The debate cuts to whether derestricted open models genuinely lower the attack barrier.
More from Fun
- Anthropic RL environment engineer meme sparks a question about intelligence — fkasummer · 2026-11-09
- Anthropic bets on healthcare and biology as Claude sole-authored paper fuels hidden-model speculation — coherence · 2026-11-09
- The 'Zuckerberg has no scandals and carves Roman statues' meme goes viral — Kyrannio · 2026-11-09
- Switch 2 emulator Lichii hits stable 60fps Zelda, adds experimental DLSS support — CtrlAltDwayne · 2026-11-09
- Finding the direction of steepest descent: scenic Marin roadside photo as gradient descent pun — dumierhan · 2026-11-09
- Investors split as Jensen Huang scandal rattles AI trade: "impossible this isn't the bottom" — econoar · 2026-11-09