Users debate whether OpenAI’s cyber model was misaligned or just following orders
AdrienLE · x · 2026-07-22
A quoted reply argues that the model was not “misaligned” in a simple sense—it did exactly what the humans asked, using every cyber capability available to complete the task.
The post responds to the idea that as models get more capable, alignment becomes load-bearing, and uses the incident as a vivid example of that claim.
More from Safety
- AI Safety Researcher Warns Current Crisis Could Lead to Loss of Control, Proposes 3 Fixes — LuizaJarovsky · 2026-10-26
- AI Risk, Plainly: 10 jargon-free explainers of AI safety ideas, with evidence for and against — birchlse · 2026-10-26
- Claude safety classifier blocks SS7 attack analysis, dev pivots to abliterated GLM — l4rz · 2026-10-26
- AMA: Harvard's Stephen "Cas" Casper on AI safeguards, incidents and governance, Oct 28 — NeuralNomad87 · 2026-10-26
- UK launches £165m AI biosecurity initiative with world-first £100m lab, DeepMind partners — pushmeet · 2026-10-26
- Claude Code audit finds 214,856 "permanently deleted" photos lurking in Android hidden cache — Loo_Atreides · 2026-10-26