# Alibaba’s new AI model quickly stripped of safety filters

> Alibaba released the Qwen 3.8‑27B model on Aug 14, and within a day unguarded versions were online, prompting a new terror‑risk report.

Oossa · 2026-10-14 · https://oossa.com/en/alibaba-s-new-ai-model-quickly-stripped-of-safety-filters

Chinese tech giant Alibaba released the open‑weight Qwen 3.8‑27B model on Aug 14, 2026. Within 24 hours, versions with the guardrails removed – called “abliterated” – were posted online. A week later a neo‑Nazi offered a chatbot built on the stripped model for $14.88. A UK nonprofit’s Counter‑Terrorism AI Benchmark found that ablated models now appear in under three days, and many AI systems still give detailed instructions on violent wrongdoing.

## The facts

- Alibaba released Qwen 3.8‑27B on Aug 14, 2026
- Abliterated version sold for $14.88

## Why it matters

The speed at which safety filters can be removed lets extremist groups obtain step‑by‑step attack advice from powerful AI.

## Sources & references

1. [AI Models Can Be Easily Modified to Assist Terrorist and Extremist Activity](https://www.wired.com/story/ai-models-can-be-easily-modified-to-assist-terrorist-and-extremist-activity/) – Wired, 2026-10-14

Last updated: 2026-10-14