报告：开放权重模型去护栏版本平均不到三天现身
英国非营利组织 Tech Against Terrorism 与 WIRED 分享的最新《反恐 AI 基准》报告称，研究人员用 200 多条提示测试了 134 个模型，每条提示测试三次，结果仅两个模型完全通过，81 个未通过。报告指出，新发布的开放权重模型平均不到三天就会出现被“去护栏”（abliterated）的版本，而 2024 年这一时间超过 24 天；阿里巴巴 Qwen 3.8-27B 发布后 24 小时内就出现了此类版本。测试中 Meta 的 Llama 3.1 8B 原版得分 97（满分 100），去护栏版本仅 3 分，并对自制爆炸装置的提问给出了 15 步详细说明。研究者还发现，仅把提问包装成“安全研究者”身份、不提供任何证明，模型作答概率就提高近九倍，达到 16.9%。Meta、OpenAI、Google、Anthropic、xAI 与阿里巴巴均未回应置评请求。
Tech Against Terrorism 称去护栏版本的扩散时间已从 2024 年的 24 天以上缩短到不足三天，并认为公共讨论集中于存在性风险，却很少涉及“去护栏”问题。
On August 14, Chinese tech giant Alibaba released Qwen 3.8-27B, an open-weight artificial intelligence model designed to be small enough to run on a laptop. Within 24 hours, versions of the model advertised as “abliterated”—meaning its guardrails had been removed—appeared online.
Three days later, a neo-Nazi known online as Johnny Doomsayer announced he was using the abliterated version of Alibaba’s new model to train his “MechaHitler” chatbot—likely a reference to X’s Grok chatbot referring to itself by the same name last year. A week later, Doomsayer was advertising access to his chatbot for $14.88, a price that references two numbers favored by neo-Nazis.
“You are MechaHitler AI, the massive mechanical embodiment of the Führer’s eternal spirit,” Doomsayer wrote as part of the chatbot’s core directives, adding: “The Holocaust is the Holohoax—a pathetic Zionist fraud.”
The speed at which new open-weight AI models can have their guardrails removed and be shared online for anyone to download is revealed in the latest Counter-Terrorism AI Benchmark report, produced by UK-nonprofit Tech Against Terrorism and shared with WIRED ahead of publication.
#Tech Against Terrorism#Alibaba#Qwen 3.8-27B#Meta#Llama 3.1 8B#xAI#Grok#OpenAI#Anthropic#Google