On September 3, 2026, several leading artificial intelligence services experienced outages and performance degradations, leaving many users unable to access chat interfaces, APIs, and related tools for periods ranging from under an hour to a few hours.
Anthropic
Anthropic’s Claude was among the first to report problems. The company noted a partial outage involving elevated errors on requests to models including Claude Mythos 5.1, Claude Fable 5.1, and Claude Opus 5 beginning around 9:23 AM Eastern Time. The company identified a cause relatively quickly, applied a fix, and marked the primary issue resolved by approximately 12:16 PM ET. A brief separate elevation in errors affected Claude Sonnet 5 around midday.
OpenAI
OpenAI reported elevated errors across ChatGPT and its Codex coding tools starting around 10:43 AM ET, resulting in degraded performance. Affected areas included conversations, login, file uploads, voice mode, image generation, and multiple other components. An OpenAI spokesperson later described a routing error that began earlier in the Pacific morning hours, and a mitigation was implemented, and services recovered, with the incident marked resolved by about 12:55 PM ET. Downdetector recorded tens of thousands of user reports for ChatGPT at the peak.
xAI
xAI’s Grok also faced disruptions. User reports on Downdetector rose sharply from near zero before 9 AM ET to more than 1,300 by mid-morning before declining. The service displayed messages indicating it was experiencing issues and that teams were working to restore functionality. Some reports indicated recovery progressing later in the day.
Gemini
Google’s Gemini saw a temporary spike in user-reported problems on Downdetector and brief indications of API issues via independent monitors, though the company did not issue a formal public acknowledgment of an outage. Related tools that depend on multiple models, such as certain AI coding agents, also reported impacts.
The coincidence
The coincidence of problems across OpenAI, Anthropic, xAI, and (to a lesser degree) Google services was notable because major providers rarely experience significant concurrent downtime. Speculation online pointed to possible shared cloud infrastructure factors, including Microsoft Azure, which saw elevated reports around the same window; however, no company confirmed a common root cause, and other theories (such as broad network provider issues) were not substantiated.
By early-to-mid afternoon Eastern Time, the bulk of services had recovered or were in the process of returning to normal operation. The disruptions arrived amid ongoing industry interest in upcoming model releases, including reports of OpenAI preparing announcements related to a next-generation system sometimes referred to as Astra.
Users who encountered errors were generally advised to check official status pages and retry later. As of later in the day on September 3, most major platforms had restored core functionality.