For a few hours on Thursday, a remarkable slice of the AI industry went dark at the same time. ChatGPT, Anthropic's Claude, xAI's Grok and OpenAI's Codex all suffered outages or elevated error rates on September 3, 2026, disrupting chatbots, coding agents and APIs that millions of people and businesses rely on daily.
The coincidental timing of the disruptions, which hit four of the most widely used AI platforms within a window of a few hours, offered a vivid reminder of how concentrated the industry's infrastructure has become. For continuing coverage of that concentration and the latest AI news shaping the market, AI Buzz Wire tracks every major provider move as it happens.
What happened at OpenAI
OpenAI's problems started first. According to BleepingComputer, the outage began at approximately 10:58 AM ET on Thursday, September 3, with users reporting errors across nearly every major ChatGPT feature.
The company's status page listed at least 15 affected ChatGPT components, including Conversations, Login, ChatGPT Work, Codex in ChatGPT Desktop, the Compliance API, Search, File uploads, Voice mode, GPTs, Image Generation, Deep Research, Agent, ChatGPT Atlas, Sites, and Connectors/Apps. OpenAI's Codex services, used by developers for automated coding, were also affected, and a bug thread on the Codex GitHub repository filled with reports as the disruption stretched on.
"We are investigating the issue for the listed services," OpenAI said on its status page, according to BleepingComputer. By Thursday afternoon the company's status dashboard showed all systems fully operational again, with ChatGPT's rolling uptime over the past several months listed at 99.62 percent.
Notably, the outage came just ahead of the expected launch of Astra, OpenAI's next major AI model, which the company has confirmed is coming soon. OpenAI has not said whether the outage was related to preparations for that launch, and BleepingComputer reported it was unlikely the issues were connected.
Anthropic's Claude degrades across its model lineup
Anthropic's trouble began early in the afternoon UTC. The company's status page shows engineers opened an investigation at 13:26 UTC on September 3 into "elevated errors on requests" affecting a wide swath of the Claude lineup: Claude Mythos 5.1, Claude Fable 5.1, Claude Mythos 5, Claude Fable 5, Claude Opus 5, Claude Opus 4.8 and Claude Opus 4.6.
The incident page lists the affected surface as claude.ai, the Claude API at api.anthropic.com, Claude Code, and Claude Cowork — in other words, essentially everything Anthropic's customers touch, from the consumer chatbot to the developer API to the coding agent many engineers keep open all day.
Anthropic identified the cause at 13:41 UTC, deployed a fix by 16:06 UTC and declared the incident fully resolved as of 9:16 AM Pacific (16:16 UTC). The episode lasted roughly three hours from first report to resolution.
Grok, Gemini and the wider ripple
xAI was hit too. A Grok outage appeared on the company's status page on September 3 and became one of the most-upvoted stories on Hacker News, where users traded reports of failed requests across OpenAI, Anthropic and xAI services in a thread asking why the providers were all down at once.
Google's Gemini was also drawn into the event, though its problems appear to have been less severe. PCMag reported that "ChatGPT, Claude, Gemini, and More Are All Down Right Now," and both the New York Post and LADbible described thousands of user complaints across the chatbots, citing outage tracker spikes. NewsNation summarized the scene bluntly: an "AI outage hits Claude, OpenAI, Grok as users report disruption."
There is no evidence the simultaneous failures resulted from a coordinated attack. Each provider described its own incident — elevated errors and component failures in OpenAI's case, an identified-and-patched error condition at Anthropic — and the near-simultaneous timing appears to be exactly what it looks like when several of the world's largest AI platforms strain under global load on the same afternoon.
When every assistant goes down at once, work stops
For consumers, a few hours of chatbot errors are an annoyance. For the growing number of businesses that have wired these systems into daily operations, the episode was closer to a stress test.
Developers who build on Claude Code or Codex lost their AI pair programmers mid-task. Customer-facing products that call OpenAI or Anthropic APIs returned errors to end users. Teams that lean on Deep Research, Agent features or ChatGPT Work for document processing simply had to wait. The Hacker News discussion filled with developers describing halted pipelines and asking whether a common dependency — a shared cloud region, a shared content-delivery layer — could be to blame. None of the providers has indicated any such common cause.
The episode echoes a familiar lesson from cloud computing, now replaying in the AI era: when a handful of providers carry most of the industry's traffic, their bad afternoons become everyone's bad afternoons. Companies that route requests across multiple model providers, cache responses where possible, and subscribe to provider status pages had a materially easier day than those with a single point of failure.
The takeaway
No provider has published a detailed root-cause analysis yet. Anthropic's incident report confirms an identified cause and a deployed fix; OpenAI's dashboard simply returned to green. If post-incident write-ups follow, they will be worth reading closely, particularly for what they reveal about capacity planning as model launches like Astra stack new demand onto already-strained systems.
For now, the September 3 outages stand as the clearest demonstration yet that AI's convenience rests on a thin, interdependent physical and software stack — one that can wobble across companies at the same time.
Stay Ahead of Every Major AI Development
Major outages, model launches and policy fights are accelerating. Get independent coverage of every development before it matters.
---
Stay Ahead of AIGet the latest AI news, analysis, and breakthroughs — all in one place.
Read more AI news →
