Key Takeaways

  • Claude, Claude Code, and Claude Cowork all went down Tuesday morning, along with sign-in systems, for close to two hours.
  • Downdetector logged more than 18,000 outage reports against a normal baseline of about 10, and Anthropic still hasn’t explained what caused it.
  • This is at least the second major Anthropic outage since August, and the pattern is starting to say more than the outages themselves.

Jump to:

I tried to log into Claude yesterday morning around 10 AM Chicago time and got nothing. Not an error message I could screenshot and laugh at. Just a spinning wheel, then a login screen that wouldn’t take my password. My first thought was I’d fat fingered something. My second thought, once half of tech Twitter started posting the same complaint at the same time, was that Anthropic had done it again.

Here’s my take. A lot of us use these tools every single day now, for research, for code, for cleaning up our own writing before it goes out into the world. So when one of them goes down for two hours on a Tuesday morning, that’s not a quirky glitch anymore. It’s a small preview of what happens when the thing you’ve built your workflow around just stops answering.

What Actually Went Down Tuesday Morning

The outage started around 10 AM ET on September 29, 2026, and it wasn’t small. Claude.ai, Claude Code, Claude Cowork, the Claude API, platform.claude.com, and the sign-in systems including single sign-on and Sign in with Apple all took a hit. Anthropic confirmed it was in what it called recovery mode after spotting elevated error rates, and at one point the company’s own advice to users was basically don’t log out, because getting back in was the hard part.

According to Android Authority’s live coverage, the worst of it ran for close to two hours before things were mostly resolved, and some messages sent during the window may not have saved at all. Downdetector logged more than 18,000 reports during the incident, compared to a normal baseline of around 10. That’s not a hiccup, that’s a full stop for a huge chunk of Anthropic’s user base at the same time.

This Was Not a One Time Thing

If this were an isolated event I’d chalk it up to bad luck and move on. It’s not. Back in August, Claude.ai, Claude Code, and Claude Cowork went down again for about 42 minutes, and Anthropic’s public explanation then was about as thin as it was this week. Its status page said “investigating” and left it there, no real root cause analysis, no promise of what changes.

Read the Bleeping Computer writeup of that August incident next to this week’s and it’s practically the same story with the date changed. Elevated errors, a status page update, a slow crawl back to normal, and then silence on what actually broke. When it’s happened this many times in a year, “we’re investigating” stops being an update and starts being a dodge.

What This Says About How Much We’ve Handed Over

I’m not anti AI. I use Claude constantly and I’m not about to pretend otherwise. But I know developers who build their entire day around Claude Code, teams that run their whole workflow through it, and regular people who use it the way they used to use Google. When that goes dark for two hours in the middle of a workday, it’s not a minor inconvenience, it’s lost work, missed deadlines, and a very uncomfortable reminder of how much control we’ve handed to one company’s servers.

That’s the bigger conversation nobody at these companies wants to have out loud. I wrote a while back about what Zuckerberg, Musk, and Altman said about how AI changes everything, and it’s easy to nod along to that kind of talk when the servers are up. It hits different when you’re staring at a spinning wheel and realizing you have no backup plan because you never thought you’d need one.

Outages happen, I get it, every tech company has bad days. What bugs me is the lack of a real explanation and the fact that this keeps happening often enough that it doesn’t feel like bad luck anymore. If you’re building your job around one of these tools, and a lot of people reading this are, it might be worth keeping at least one manual process alive for the days it goes dark. Because it will go dark again.

Frequently Asked Questions

What happened during the September 2026 Claude outage?

Claude.ai, Claude Code, Claude Cowork, the Claude API, and sign-in systems all experienced elevated error rates starting around 10 AM ET on September 29, 2026. Anthropic said it entered recovery mode and mitigated the initial impact, though problems with signing in persisted afterward.

How long was Claude down?

The worst of the outage lasted close to two hours, with Downdetector logging more than 18,000 reports compared to a normal baseline of around 10.

Is this the first time Claude has gone down in 2026?

No. Anthropic confirmed a separate outage in August 2026 that lasted about 42 minutes and affected the same core services. Neither incident came with a public root cause explanation.

What should I do if Claude goes down again?

Save your work elsewhere as you go if you’re relying on Claude for anything time sensitive, and keep a fallback process in mind for coding or writing tasks so a two hour outage doesn’t turn into a lost afternoon.