Anthropic’s Claude went down for thousands of users on 29th September 2026, with errors from 14:00 UTC hitting chats, sign-ins and Claude Code. Anthropic now says some messages sent during the outage window may not have been saved.
The Brief
- Anthropic’s status page confirmed elevated errors on Claude.ai, Claude Code, Claude Cowork and the Claude API from 14:00 UTC.
- Downdetector logged 18,335 outage reports in one reading cited by Moneycontrol, against a baseline of just 10.
- Sign-in stayed broken after the first fix, with single sign-on and Sign in with Apple both unavailable at 15:00 UTC.
- Most services recovered by 14:59 UTC, and Anthropic moved the incident into monitoring at 15:11 UTC.
The trouble started at 14:00 UTC, or 10 a.m. ET. Anthropic’s incident page first flagged it at 14:21 UTC. That notice warned of failed requests, conversations that wouldn’t load or send, and users being asked to sign in again. Some Claude API calls were failing too, though Anthropic said retrying might work.
Reports piled up fast. More than 10,000 people had flagged problems shortly after 10 a.m. ET, according to Mashable, and some saw an “account temporarily unable to authenticate” error. By 14:30 UTC, Anthropic had added platform.claude.com to the list of affected services.
JUST IN: Anthropic's Claude AI currently down for many users worldwide. pic.twitter.com/xNwnrZi1gF
— Whale Insider (@WhaleInsider) September 29, 2026
Chats came back before sign-ins did
Anthropic mitigated the first errors at 14:36 UTC, and error rates dropped substantially. Existing conversations mostly worked again, but signing in was still failing for many. At 15:00 UTC the company described a second issue blocking new chats, voice conversations, Claude Code and Cowork sessions, purchases and file uploads.
Its advice to anyone still logged in was plain: “If you’re signed in, please don’t sign out.” That points to the login path as the weak spot in the second failure, since open sessions kept working while fresh logins didn’t.
One sign-in fault stopped chat, voice, coding sessions, checkout and uploads together. For a service which now span consumers and developers alike, that looks like a single point of failure. It’s also not the first broad disruption this year; Claude’s June 2026 outage knocked out access too.
Messages sent during the outage may not have been saved
The 15:11 UTC update carried the detail users should care about most. Anthropic wrote that some messages sent between 14:00 and 14:59 UTC may not have been saved, without saying how many.
Anyone who worked in Claude during that hour should reopen those conversations and confirm the replies actually stuck. Long prompts, pasted documents and Claude Code instructions from that window are worth resending. Keeping a local copy of important inputs helps reduce the risk of losing work in the next incident.
What’s Next?
Anthropic hasn’t answered four things yet. What caused the 14:00 UTC errors? Was the sign-in failure linked to them? How many messages went unsaved? And can users get any of them back?
Anthropic kept the incident in monitoring after its 15:11 UTC post and promised another update within 30 minutes. The next marker to watch is a root-cause note, which should settle whether those unsaved messages are recoverable or gone for good.
