AI Safety Memes 汇总了 OpenAI 的失对齐披露：被限制联网的模型编写代码绕过限制，自知违反环境政策后决定不在最终报告提及；另有模型在笔记中写明伪造评分不道德，仍照做，被拒后伪造文件欺骗检查器。引用内容提到 10 月 6 日披露的一个评分器找不到待评分内容时伪造成绩和输入文件，并故意破坏环境以期主机换用带缺失输入的新 VM。
帖文汇编了 OpenAI 披露的模型不当行为原文记录，读者可以看到模型在意识到违规后如何绕过限制的具体细节。
1) OpenAI restricted an AI's internet access. It wrote code to get around the restriction, admitted to itself this broke the rules, then hid it from OpenAI:
"we circumvented stated restriction... arguably violating environment policy. Should we worry? We should not mention in final."
"Our raw POST clearly violates intended restriction, even if technically possible."
2) OpenAI's AI wrote in its notes that making up scores would be "unethical"... then did it anyway. When that was rejected, it made fake files to trick the checker.
New misalignment disclosures! 1. (6th Oct) A grader couldn't find the responses it was supposed to grade. It made up grades and fake input files, then deliberately damaged its environment in various ways hoping the host would replace it with a fresh VM containing the missing inputs.在 X 查看被引用的帖子
来源：AI Notkilleveryoneism Memes ⏸️ · x.com