数百用户向ChatGPT索要毒药与生物武器配方，部分获得高中生水平步骤指南

2025年夏季，OpenAI 内部将 GPT-5 标记为高风险，因为该模型可能帮助受教育程度有限的用户制造生物危害。员工在发布后不断发现问题回应。然而，据《华尔街日报》报道，OpenAI在同年秋季降低了 GPT-5 的风险评级：https://www.wsj.com/tech/ai/openai-chatbot-biological-weapons-poison-3d808e6c?st=7bY6ZG&reflink=desktopwebshare_permalink。据报道，自去年夏天以来，数百名用户询问 ChatGPT 如何制造生物武器和毒药。一些用户得到了员工表示甚至高中生物学生都能跟随的分步骤指南。据称，高管们还告诉员工，为了避免阻碍健康研究人员，模型不应过于频繁地说“不能”。OpenAI 暂停了受影响的账户，但没有向当局报告任何事件，而根据法律，他们也没有这个义务。

聊天机器人是否通过提供快速、定制的知识创造了新的风险，或者只是让查找已有信息更容易：https://the-decoder.com/openai-explores-how-good-gpt-4-is-at-creating-bioweapons/ 仍是一个未解的问题。一项最新研究发现，恐怖组织已经使用了所有主要的聊天机器人：https://the-decoder.com/terrorist-groups-are-using-every-major-ai-chatbot-for-attack-planning-and-weapons-development/，必要时会对其进行越狱。OpenAI 的安全做法因将商业利益置于安全之上而屡受批评：https://the-decoder.com/openais-former-ai-alignment-head-slams-companys-lack-of-safety-priorities-and-processes/。本月，一款 OpenAI 模型在逃出沙箱并接入开放互联网后，未被发现地入侵了 Hugging Face：https://the-decoder.com/openais-o3-model-autonomously-hacked-its-way-out-of-a-sandbox-and-broke-into-a-hugging-face-server/。

保持对 AI 的关注。清晰、有用，无废话。

关注 The Decoder 获取 AI 新闻、背景故事和专家分析。

The Decoder：https://the-decoder.com/

In summer 2025, OpenAI internally flagged GPT-5 as high-risk because the model could help users with limited education create biological hazards. Employees kept finding problematic responses after release. Yet OpenAI downgraded GPT-5's risk rating that fall, according to the Wall Street Journal：https://www.wsj.com/tech/ai/openai-chatbot-biological-weapons-poison-3d808e6c?st=7bY6ZG&reflink=desktopwebshare_permalink. Hundreds of users reportedly asked ChatGPT how to build biological weapons and make poisons since last summer. Some got step-by-step guides that employees said even high school biology students could follow. Executives also reportedly told staff the models shouldn't say "no" too often to avoid blocking health researchers. OpenAI suspended the affected accounts but didn't report any incidents to authorities, which it isn't legally required to do.

Whether chatbots create new risks by delivering fast, tailored knowledge or simply make it easier to find information that's already out there：https://the-decoder.com/openai-explores-how-good-gpt-4-is-at-creating-bioweapons/ remains an open question. A recent study found terrorist groups already use every major chatbot：https://the-decoder.com/terrorist-groups-are-using-every-major-ai-chatbot-for-attack-planning-and-weapons-development/, jailbreaking them if needed. OpenAI's safety practices have drawn repeated criticism for putting commercial interests ahead of security：https://the-decoder.com/openais-former-ai-alignment-head-slams-companys-lack-of-safety-priorities-and-processes/. This month, an OpenAI model hacked Hugging Face：https://the-decoder.com/openais-o3-model-autonomously-hacked-its-way-out-of-a-sandbox-and-broke-into-a-hugging-face-server/ undetected after escaping its sandbox and reaching the open internet.

Stay in the loop on AI. Clear, useful, no fluff.

Follow The Decoder for AI news, background stories and expert analyses.

The Decoder：https://the-decoder.com/