Abliterated GLM-5.3 released to run offensive cyber and red-team work other models refuse
thedealdirector · x · 2026-09-03
The abliterationai account released abliterated-model-large-v2, based on GLM-5.3 (ranked #3 on Terminal-Bench 4.0, behind only Opus 5 and Fable), claiming 2× the cyber exploitation capability of 5.2. The model was abliterated and hosted to perform offensive cyber, red-teaming, and agent testing that other models refuse. Specs: US-hosted, FP8, 1M context window, zero input/output prompt retention. A notable AI-safety event around publicly available de-aligned models.
Related event: Abliterated GLM-5.3 Model Release Sparks Cybersecurity Policy Concerns(9 posts)→
More from Safety
- Prompt injection via email can hijack your AI agent — jev-guard adds risk-scored guards — sujingshen · 2026-10-18
- Users press OpenAI on long-delayed Adult Mode a year after it was promised — ___Patrice___ · 2026-10-18
- Viral AI safety report debunked: 'stolen keys' were just a 'test' string — basedjensen · 2026-10-18
- One diagram explains the difference between CSRF and XSS attacks — _jaydeepkarale · 2026-10-18
- Grady Booch slams Anthropic for tolerating up to 50 deaths and $1B damages per model — Grady_Booch · 2026-10-18
- Single Neuron Suffices to Bypass Safety Alignment in LLMs, Paper Finds — burny_tech · 2026-10-18