ChatGPT, Claude, and Grok all went down at the same time

Key points

• ChatGPT began returning errors at around 11AM ET, with OpenAI confirming "elevated errors" across ChatGPT and its Codex coding tool.
• Anthropic's Claude started showing problems around 10:30AM ET, affecting five model versions before narrowing to two: Opus 4.8 and Opus 5.
• XAI's Grok went down across its app and website at roughly the same time.
• No cause has been confirmed, and no company has said whether the three outages share a common origin.

Three of the most widely used AI chatbots stopped working properly at almost exactly the same time on Tuesday morning, leaving millions of people unable to get responses from tools many now rely on for work.

ChatGPT, made by OpenAI, started throwing error messages at around 11AM ET. OpenAI's status page flagged "elevated errors across ChatGPT and Codex" (Codex is OpenAI's tool for writing and checking code). OpenAI has applied some form of fix and says it's watching whether the service recovers.

What happened to Claude?

Anthropic's Claude began having problems around 10:30AM ET. Five versions were initially affected, among them Opus 4.8 and Opus 5. By publication, the trouble had narrowed to those two versions alone, with Anthropic saying it's "continuing to work on a fix."

Grok, the chatbot from Elon Musk's xAI, was also down across its mobile app and website at roughly the same time. The company said it's "working on restoring service as quickly as possible" and offered no further detail.

Are the three outages connected?

Nobody knows yet. Three separate companies, three separate products, all going dark within a thirty-minute window: the timing is hard to ignore. But coincidence hasn't been ruled out, and so has the possibility of a shared piece of internet infrastructure that all three services depend on. The Verge AI, which first reported the story, said it had not heard back from any of the three firms at the time of publication.

We've covered Grok and its parent company closely in recent weeks. Our 27 August story on a federal lawsuit alleging xAI trained Grok on child sexual abuse images drew significant reader attention, and a fortnight later the Pentagon deployed custom versions of both ChatGPT and Grok to three million staff. Simultaneous outages across all three of the major chatbots those users might reach for will not be a comfortable footnote to either story.

For anyone whose work depends on a single AI service, today's events are a plain reminder: these platforms go offline without warning, and a backup plan isn't optional.

This story is still developing. We'll update it when any of the three companies explains what happened.