ChatGPT, Claude and Grok failed within the same window. OpenAI's status page reported elevated error rates in ChatGPT and Codex, and the company said it had already applied a fix and was watching the recovery. Anthropic's status page said the problem by then reached only Opus 4.8 and Opus 5, with its other models back to normal error levels. Grok was seeing large-scale failures on X. One user on X summed the whole thing up as every AI service going out at once.
On Downdetector the curve spiked and then dropped, which indicates OpenAI's outage was largely contained. Other services were still broken while it recovered, including Google and Fortnite.
The reaction was immediate and mostly unserious. Users turned a rare simultaneous failure into jokes about a society leaning on technology that is supposed to be eroding its ability to think unaided. One wrote that he could finally see the sun. Paris Marx, a journalist and well-known critic of the industry, joked that for a short stretch millions of people had to use their own brains again. Another user proposed a thought experiment about what happens during a global outage to someone who rents his brain from an AI.
The jokes are the interesting data point, because they describe a dependency that did not exist to be joked about a short while ago. A chat product going down used to be an inconvenience for people mid-draft. The same outage now takes Codex with it, which means it takes working software engineers offline, and it arrives alongside a vendor status page listing which specific model versions are and are not answering.
What none of the three companies said is why. The status pages name symptoms — elevated error rates, affected models, a fix applied — and stop there. Three vendors who compete with each other, plus Google and a video game, degrading in the same window is the kind of coincidence that usually has a shared cause underneath it, and my read is that the missing explanation, not the downtime, is the story here. Nobody is obliged to publish a root cause on the day. Nobody did.
That gap matters more as the products move from answering questions to taking actions. An assistant that writes your code, books your travel and moves money on your behalf is infrastructure, and infrastructure gets judged on the incidents it explains, not the ones it recovers from quickly. Right now the industry is selling the first thing and reporting like the second.