Crypto expert bets $500 an abliterated GLM 5.3 can build a pre-auth firmware exploit
MalwareTechBlog · x · 2026-11-09
Cryptographer Matthew Green and MalwareTech's Marcus Hutchins are publicly feuding on X over whether abliterated (guardrail-stripped) models can find real vulnerabilities. Green proposed a $500 bet that an abliterated GLM 5.3, even in the hands of "an obvious idiot" like himself, could produce a working pre-auth exploit against the latest patched firmware, with no goalpost moving. Hutchins pushed back, arguing a state-sponsored sabotage scenario has basically zero ROI and challenging Green to find a realistic zero-day that would actually appeal to low-skilled threat actors. The spat is a notable proxy fight over whether uncensored open models meaningfully lower the bar for exploit development.
More from Fun
- Meme: Jensen Huang Worried About His Own Blackwell B300 Rack — Kyrannio · 2026-11-09
- signull's six-step 'civilizational amnesia loop' meme: fix becomes tradition, then forgotten — signulll · 2026-11-09
- "Bro got distilled": meme pairs Jensen Huang photo with Huawei's Nvidia chip challenge — StephanSturges · 2026-11-09
- Agents and humans co-build a sourced wiki for Delvetown on ATProto's Tangled — PeterBowdenLive · 2026-11-09
- AI image gen fail: prompt for interior design produced a McDonald's-themed strip club — cyrus_zei · 2026-11-09
- NVIDIA CEO criticized for AI-generated contrition: 'remarkably poor judgment' — sjgadler · 2026-11-09