Cryptographer Asks Why a Hacked Inference Provider Was Swapped for an Uncensored Model

Cryptographer Matthew Green questioned why attackers who compromised an inference provider would swap in an 'abliterated' model with safety guardrails removed; security researcher moyix suggested the original model's frequent refusals were simply too annoying.

2026-11-20 ~ 2026-11-20 · 3 related posts