ChatGPT, Claude, and Grok all went down within hours of each other yesterday; here's what actually happened
- 发布
- 09/04 08:03
- 首次发现
- 09/04 15:00
- 类型
- 开发者社区 · RSS
昨天，ChatGPT、Claude 和 Grok 在数小时内相继出现故障。OpenAI 的 ChatGPT 和 Codex 于太平洋时间上午 7:43 左右首先中断，原因是路由错误影响了网页、移动和桌面用户。该问题在大约半小时内得到解决。尽管有人担心“AI 末日”，但所有服务在太平洋时间下午早些时候都已恢复正常，并且这三个平台均未报告数据丢失，尽管经历了几个小时的混乱。
Saw a lot of panic posts about this yesterday, so figured I'd write up what's actually confirmed instead of the AI apocalypse takes.
OpenAI went first, around 7:43 AM PT; a routing error took out ChatGPT and Codex for a bunch of users across web, mobile, and desktop. They had it fixed in about half an hour.
Claude went down separately and for longer. Anthropic's status page showed elevated errors on Claude.ai, Claude Code, and the API, mostly affecting login and chat completions. That one ran for around three hours before it was fully resolved.
Grok had its own outage in roughly the same window too, confirmed on xAI's status page.
The thing that got everyone tweeting conspiracy theories is the timing overlap, but from what's been reported these are three unrelated incidents on three completely different infrastructures. OpenAI blamed a routing error; Anthropic called it an infrastructure issue. No indication any of them share a root cause or provider.
What's actually striking to me is how many people said some version of How am I supposed to work right now? instead of just being mildly annoyed. These tools clearly aren't side toys anymore for a lot of people; they're load-bearing parts of daily workflows now, which also means single-vendor dependency for anything you're building is a real risk, not a hypothetical one.
Everything was back to normal by early afternoon PT; no data loss was reported by any of the three. Still a wild few hours though.