ChatGPT, Claude and Grok Go Down Together Thursday, Downdetector Logs Over 36,000 Reports

Three of the world's biggest AI chatbots stopped working for thousands of people at the same time Thursday morning, and by midday nobody had a straight answer why.

Downdetector showed reports for OpenAI's ChatGPT topping 36,700 by 11 a.m. ET, according to News4Jax. About 80% of those complaints were tied to the core ChatGPT product, with smaller shares hitting Codex and the OpenAI app. PCMag put the peak even higher, over 36,000 reports, and said the OpenAI trouble ramped up fast starting around 10:30 a.m. ET.

Anthropic's Claude had problems earlier. PCMag reported a small spike around 8:30 a.m. ET tied to Claude Sonnet 5, then a bigger wave an hour later that hit more models, including Opus 5 and what PCMag called Mythos/Fable 5.1. Anthropic's own status page said Opus 4.8 and Opus 5 remained the stubborn holdouts even after a fix rolled out, according to Axios.

X's Grok was down too. xAI confirmed the service issue, according to 9to5Mac. Cursor, a coding tool that leans on both Claude and Grok, went down as a direct result, 9to5Google reported.

Gemini's status is murkier

Google's Gemini saw its own spike in Downdetector reports, per both PCMag and Axios, but Google never posted an official outage confirmation. 9to5Google said Gemini "appears to be unaffected," while PCMag reported a staffer got error messages on every prompt they tried. A real but smaller-scale disruption could hit without crossing Google's threshold for a public acknowledgment.

What might have caused it

Axios and 9to5Google both flagged that Microsoft Azure, which provides cloud infrastructure for ChatGPT, Claude and Grok, was seeing its own spike in outage reports at the same time. That's a plausible shared point of failure, since all three platforms lean on Azure. PCMag raised a second theory, pointing to a simultaneous outage at Cloudflare, though it stopped short of calling either one the confirmed cause.

None of the companies involved, OpenAI, Anthropic, xAI or Microsoft, had publicly identified a root cause as of the reports gathered Thursday. OpenAI's status page said only that a fix had been "applied" and it was "monitoring the recovery," per Newsweek.

Mathrubbumi English noted that users on X speculated traffic fleeing a down Claude overloaded ChatGPT, but there's no verified evidence that's what happened. High traffic, a bad software push, a shared vendor problem, or a regional network failure could each explain it independently.

Recovery, sort of

By roughly 8:49 a.m. PT, 9to5Google reported OpenAI said a fix had been applied and ChatGPT was recovering. Claude said most models were back except Opus 4.8 and Opus 5. Grok's dashboard still just showed an outage with no update. 9to5Mac reported service "coming back online for some" about half an hour after it went down, with the caveat that results varied by user.

The timing added a wrinkle. 9to5Mac noted OpenAI is rumored to be close to launching a new model, code-named Astra, that could bump ChatGPT from GPT-5.6 to GPT-6. That's an unconfirmed rumor, not a fact, and no source tied the outage to any model rollout. Whether the two things are connected, or just landed in the same week by coincidence, is an open question OpenAI hasn't addressed.

A huge share of the country's coding, writing, research and customer service work now runs through a small handful of AI companies, all leaning on the same handful of cloud providers. When Azure or Cloudflare sneezes, apparently, everybody catches a cold. Whether any of these companies will publish a real post-mortem explaining what broke, and whether it points back to a single shared vendor, remains to be seen.