Widespread outage hits ChatGPT, Claude and Grok

What happened

On Thursday, September 3, 2026, several major AI tools experienced simultaneous disruptions. Users worldwide reported problems with OpenAI's ChatGPT, Anthropic's Claude, and X's Grok. The outages were confirmed both by a spike in reports on DownDetector and by the companies' own status dashboards. According to the Asbury Park Press, DownDetector logged more than 37,000 reports for OpenAI/ChatGPT, over 1,300 for Claude, and more than 1,000 for Grok on Thursday morning. Reports for Google's Gemini also rose, exceeding 300, though 9to5Google notes that Google did not officially confirm an outage and Gemini appeared to keep working normally. The AI coding tool Cursor also went down, attributing its own outage directly to the Grok and Claude disruptions.

Possible cause

9to5Google suggests the simultaneous failures might be linked to problems at Microsoft Azure, which also saw a spike in outage reports at the same time. OpenAI, Anthropic, and X all partly rely on Azure's cloud infrastructure, although each also uses other providers. At the time of publication, no official cause had been confirmed.

OpenAI confirms a separate incident

German outlet PlayCentral reports that OpenAI logged a new, distinct incident on its official status page on the afternoon of September 3, around 16:58, titled "Elevated errors across ChatGPT and Codex." The status was listed as "Investigating," and the problem was classified as degraded performance rather than a full outage, affecting 15 ChatGPT components and 4 Codex components. PlayCentral stresses this is a separate incident with its own ID, unrelated to the "Work mode" disruption from August 31. The outlet's own measurement showed ChatGPT's website remained reachable, with a response time of around 73 milliseconds, though individual features such as answer generation could still be affected by elevated error rates.

Advice for users

PlayCentral recommends checking whether a normal conversation works outside of code or work mode, reloading the page, testing the app and browser separately, temporarily switching to another available model, and saving longer inputs locally before resubmitting them. Since this is a server-side issue at OpenAI, changes on the user's end typically do not help — waiting for a fix remains the main option.

Separate report of verification codes

German blog Caschys Blog (published via stadt-bremerhaven.de) describes a seemingly unrelated issue: several users in Germany report repeatedly receiving temporary verification codes from OpenAI's official address noreply@tm.openai.com, despite not having a ChatGPT account. The blog's author speculates that leaked email addresses may be used to attempt account registrations on ChatGPT. This information comes from a single source and has not been officially confirmed.