Skip to content
AI news · Claude

Claude took an hour off on 29 September, and took a few messages with it

By , Founder of Adaptd · · Updated

An outage hit claude.ai, the Claude apps, Claude Code, Cowork and the API for about an hour. Some messages sent during it may not have been saved. A good excuse to write the plan B you've been meaning to.

The short version

  • Claude had an outage on 29 September from 14:00 to 14:59 UTC, affecting claude.ai, the apps, Claude Code, Cowork and the API.
  • Anthropic says some messages sent during that hour may not have been saved, and API automations may have failed without retrying.
  • Write a plan B now: name a second tool, keep key prompts in a shared document, and make API automations retry or alert.

If you were asleep, you missed it. If you were working late, you probably noticed, and possibly said a few words you wouldn't put in a client email.

What happened in the Claude outage on 29 September?

Claude had an outage on 29 September from 14:00 to 14:59 UTC, which was midnight to 1am on 30 September in Melbourne. According to Anthropic's status page, it affected claude.ai, the desktop and mobile apps, Claude Code, Claude Cowork and the Claude API. People saw failed requests and errors loading or sending conversations, and some were asked to sign in again.

It happened in two stages. Anthropic mitigated the first errors at 14:36 UTC, and existing conversations mostly started working again. A second problem then stopped many people signing in, because single sign-on and Sign in with Apple were unavailable. It also blocked new chats, voice conversations, Claude Code and Cowork sessions, purchases and file uploads. Anthropic told users who were still signed in not to sign out. Everything was working again by 14:59 UTC. The status page also lists a separate incident on 22 September that caused errors across several models.

Why do the lost messages matter more than the downtime?

Anthropic says some messages sent to Claude during that hour may not have been saved, and for a business that is the bigger problem, more than the hour of downtime. If your team drafts a proposal or a client reply in a chat, the chat is often the only copy. Chat history is a whiteboard: great for thinking, and nobody should be keeping the only copy of the contract on it. And if an automation calls the Claude API, it probably failed during that hour and may not have tried again.

What should your team do when its main AI tool goes down?

Write down what your team does when its main AI tool stops working. Name a second tool people can switch to, and keep important prompts and instructions in a shared document instead of in chat history. Save any output you plan to use as soon as it's written. If you have automations that use the API, check that they retry or alert someone when a call fails.

Every AI service goes down sometimes, the good ones included. It's like a blackout: you only find out where the torch is when you need it. Write the plan B while the lights are on.

Wondering what this means for your business?

That's exactly what the discovery call is for.