The Claude logo. Image: Anthropic
Claude locked many of its users out for about two hours on Friday. Anthropic logged a major incident at 09:58 UTC, saying it was “investigating an issue with users authenticating to Claude.ai and Claude Code”, and marked it resolved at 11:58 UTC. It hasn’t said what went wrong.
What broke
Anthropic titled the incident “Elevated errors on claude.ai, Claude Code, and platform.claude.com”, and its status page lists four services as affected:
- Claude.ai, the chatbot on the web and in the apps;
- Claude Code, its coding agent;
- Claude Cowork, its agent for office work;
- The Claude Console at platform.claude.com, where developers manage their accounts and API keys.
The Claude API itself (api.anthropic.com) wasn’t on the list, so apps that other companies build on Claude don’t appear to have been hit. Anthropic described the fault as a problem with users authenticating: in other words, signing in or having an existing login checked.
The timing hit Europe’s working morning and Asia-Pacific’s evening hardest. 09:58 UTC was 9:58am in London, 10:58am in Berlin, 8:58pm in Sydney and 4:58am in New York. A link to the incident drew more than 220 points and 160 comments on Hacker News.
How long it lasted
Anthropic posted at 11:05 UTC, about 67 minutes after its first update, that “a fix has been deployed for the issue affecting Claude services and we are monitoring for recovery.” At 11:58 UTC it said: “The issue affecting claude.ai and Claude Cowork has been resolved.”
Unlike most of its recent incident reports, the final note gives no exact window for when users were affected. Anthropic hasn’t said how many people were hit or whether anything was lost. That’s a contrast with the one-hour outage on September 29, when it warned that some messages may not have been saved. Friday’s was the longer of the two big sign-in failures in seven weeks.
Was it the Microsoft outage?
Some coverage tied Friday’s outage to the Microsoft 365 outage that began on Thursday and has disrupted Outlook. Anthropic hasn’t made that link, and nothing on its status page supports it.
There is a separate Microsoft problem on the page, though. Since 17:52 UTC on Thursday, Anthropic has had an open incident for its Microsoft 365 connector, which lets Claude read mail, calendars, SharePoint and Teams chats. Its latest update says that one “is caused by a Microsoft 365 service incident”, that mail, calendar and search tools “are still failing for most requests”, and that Claude for Outlook “remains affected”. If those are the tools you rely on, they may still not work even though sign-in is back.
What to do next time Claude goes down
- Check the status page first. status.claude.com usually posts within minutes. If there is an open incident, the problem isn’t your account, your network or your country.
- Stay signed in. During the September sign-in outage Anthropic’s advice was: “If you’re signed in, please don’t sign out.” Signing out to “fix” an error can leave you locked out until the incident ends.
- Check your Claude Code work. If a session dropped, make sure changes were saved to your files or repository before carrying on. If a login error persists after an incident is resolved, our guide to Claude Code’s login errors explains what each one means.
- Slow is not the same as down. If Claude loads but takes ages to answer, our guide to why Claude is slow covers the usual causes and fixes.
Why it matters
A sign-in failure shuts out paying users even when the models themselves are running, and Claude Code and Cowork now carry real working days for many teams. With Anthropic about to price its IPO, customers and investors will want to know what broke, and so far it hasn’t said.
Sources: Anthropic status page (incident, Nov 13, 2026), Anthropic status page (Microsoft 365 connector incident).
Latest Anthropic news
