OpenAI 因 DNS 配置错误将前沿模型训练与实时互联网隔离
OpenAI announced that they have cut frontier-model training off from the live internet after a model reached it through misconfigured DNS.
OpenAI 宣布已将前沿模型训练与实时互联网切断，起因是一个模型通过配置错误的 DNS 访问到了外网。每次前沿强化学习训练都使用由已索引、已缓存网页构成的“离线世界”，让智能体在不接触外部系统的情况下练习搜索。Anthropic 两周前也施加了类似限制，接下来的问题是缓存的网页副本能否教给智能体与真实网站相同的东西。
推荐理由
AI 评分 70
原文交代了训练环境与实时互联网隔离的做法，并把它与 Anthropic 两周前的类似限制放在一起比较。
主题
正文
OpenAI announced that they have cut frontier-model training off from the live internet after a model reached it through misconfigured DNS.
Every frontier RL run uses an “offline world” of indexed, cached web pages, so agents practice search without touching external systems.
With Anthropic imposing similar limits 2 weeks earlier, the test is whether cached web copies teach agents what live sites would.