Cryptographer Matthew Green prods abliterated-model inference providers on security
matthew_d_green · x · 2026-10-25
Renowned cryptographer Matthew Green quipped that he hopes inference providers hosting "abliterated" open-weight models—ones with their safety alignment stripped out—have hardened their systems. The half-joking remark highlights the real abuse and security risks providers take on when they commercialize uncensored models.
More from Fun
- Perplexity maps all 92 places Johnny Cash sings in "I've Been Everywhere" — AravSrinivas · 2026-10-25
- Kalshi traders' photo goes viral as X jokes about prediction-market gambling — kyliebytes · 2026-10-25
- Pikin Days: a free browser game recreating nostalgic Nigerian childhood errands — emeka_boris · 2026-10-25
- ChatGPT Map View Shows Direction Exactly 180° Off From Actual Compass Reading — ikeadrift · 2026-10-25
- Training quirk: reverse-chronological data ordering worsens loss — kalomaze · 2026-10-25
- Leaving the AI bubble: normal people's take on AI is basically 'GPT slaps' — djcows · 2026-10-25