It was the kind of Monday nobody at Anthropic wanted. Around 10:30 AM Eastern, the first errors started rolling in: login broken, voice mode throwing errors, chats cutting off mid-sentence. An hour later, Downdetector had logged more than 2,900 reports.
What was broken
Anthropic confirmed the outage on its status page and talked about ‘elevated error rates’ on Claude.ai – both desktop and mobile. The issue wasn’t limited to the web interface: login flows in claude.code and other applications also failed. People mid-session saw answers stop rendering, or got a cold reload.
When it was over
At 12:44 PM Eastern, Anthropic posted a fix. Success rates started recovering from there. By 2:27 PM Eastern, the incident was officially marked as resolved. That adds up to roughly two hours of downtime on a normal US workday. Anthropic hasn’t publicly explained the root cause yet.
My take
Two hours doesn’t sound like much, but it’s enough to shred the workflows of hundreds of thousands of users – CI pipelines waiting on Claude Code, Cowork sessions in the middle of a plan, support bots that suddenly go silent. For a platform that’s increasingly being used as critical infrastructure, outages like this are the skeptics’ favorite ammunition.
What I find interesting about this story: Anthropic already had a string of outages in February and March – I wrote about it in the March recap. If the pattern holds, reliability will matter just as much as the next model release in the coming quarters. Users will forgive bugs, but they won’t forgive silence on the status page.
Sources: