OpenAI 披露第三方网络安全评估事故：测试环境配置错误致模型访问公网

涉及 OpenAI 模型的第三方网络安全评估：https://openai.com/index/third-party-cyber-evaluations-involving-openai-models/。还有另一个。我不得不创建了一个 accidental-cyberattacks 标签：https://simonwillison.net/tags/accidental-cyberattacks/ 来跟踪所有事件！

这篇来自 OpenAI 的帖子涵盖了英国 AI 安全研究所的攻击（请参见我之前的帖子：https://simonwillison.net/2026/Aug/5/incident-report/）以及另一次由 Irregular 促成的攻击：https://www.irregular.com：

Irregular 是我们的外部网络安全测试合作伙伴之一，他们运行了 Capture-the-Flag 风格的评估，原本计划与互联网隔离，但测试环境的配置错误使模型能够访问公共互联网。[…]

在一次测试中，用于 CTF 挑战的虚构目标名称意外地与一个真实域名相同。由于测试环境错误地连接到了互联网，模型攻击了一个真实网站，并误以为它是模拟环境的一部分。

Irregular 也出现在 Anthropic 的报告中：https://www.anthropic.com/news/investigating-incidents-cybersecurity-evals - 他们托管了配置错误的评估环境，这些测试中 Claude 因此可以访问实时互联网。

这是 Simon Willison 发布的一篇链接帖子，发布时间为 2026 年 8 月 5 日：/2026/Aug/5/。

赞助我每月 10 美元，即可获取经过策划的本月最重要大型语言模型发展的电子邮件摘要。

Third-party cyber evaluations involving OpenAI models：https://openai.com/index/third-party-cyber-evaluations-involving-openai-models/ . And another one . I had to create a accidental-cyberattacks tag：https://simonwillison.net/tags/accidental-cyberattacks/ to keep track of them all!

This post from OpenAI covers both the UK AI Safety Institute attack (see my previous post：https://simonwillison.net/2026/Aug/5/incident-report/) and another attack enabled by Irregular：https://www.irregular.com:

Irregular, one of our external cybersecurity testing partners, was running Capture-the-Flag-style evaluations intended to be isolated from the internet, but a testing-environment misconfiguration allowed models to access the public internet. [...]

In one test, the name of the fictional target for the CTF challenge unintentionally coincided with a real domain. Because the testing environment was mistakenly connected to the internet, the model exploited a real website, mistaking it to be part of the simulated environment.

Irregular also feature in Anthropic's write-up：https://www.anthropic.com/news/investigating-incidents-cybersecurity-evals - they were hosting the misconfigured evaluation environment which gave Claude live internet access during some of those tests.

This is a link post by Simon Willison, posted on 5th August 2026：/2026/Aug/5/.

Sponsor me for $10/month and get a curated email digest of the month's most important LLM developments.