Claude outage: an hour of messages may not have saved
Today's Claude outage hit claude.ai, the API, Claude Code, and Cowork, and some messages from 14:00-14:59 UTC may not have saved. Log your own transcripts.
This morning's Claude outage took down more than chat. For about two hours, claude.ai, the Claude API, Claude Code, and Claude Cowork returned elevated errors, and sign-ins failed. That part is routine. The line that should get your attention is on Anthropic's own status page: some messages sent between 14:00 and 14:59 UTC may not have been saved.
What actually happened
Per the Claude status page, Anthropic began investigating elevated errors at 14:21 UTC (7:21 AM in Phoenix) across claude.ai on web, desktop, and mobile, api.anthropic.com, Claude Code, Claude Cowork, and authentication. A mitigation at 14:41 UTC cut the error rate substantially. At 15:00 UTC a second problem with sign-ins was still open. By 15:11 UTC most services had recovered, and Anthropic posted the note about unsaved messages. The incident was marked resolved at 16:27 UTC.
Nothing leaked. Nobody was breached. But for one hour, the vendor's record of your conversations may have gaps.
Why the Claude outage matters for your business
Downtime you can plan for. We've written before about timeouts, queues, and a second route for when a model API stops answering. Lost state is a different problem. If your team uses a chat app as the system of record, a lost hour is lost work: a support thread, a drafted contract clause, or the reason an agent did what it did.
The fix is simple and cheap. The vendor's history is a convenience, not your log.
- API integrations: write every request and response to your own store before you act on it. A Postgres table with the prompt, the output, the model, and a timestamp is enough.
- Agents and Claude Code: commit early and often. Keep task state in your repo or database, not only in the session.
- Customer-facing bots: save the transcript on your side. If a customer disputes what your bot told them, "the vendor lost it" is not an answer.
Then check your own 14:00-15:00 UTC window from today. If a job ran then, confirm the output exists where you think it does.
Key takeaways
- claude.ai, the Claude API, Claude Code, Claude Cowork, and sign-in saw elevated errors starting 14:21 UTC on September 29, 2026
- Most services recovered by 15:11 UTC; Anthropic marked the incident resolved at 16:27 UTC
- Anthropic says some messages sent between 14:00 and 14:59 UTC may not have been saved
- For your business: log every model request and response to your own store, and audit any jobs that ran in that window
If your AI's memory lives only on the vendor's servers, you don't own it. We build AI integrations that write their own audit log, so an outage costs you minutes, not records. See how we build AI into products or have us review your integration.
Sources: Claude Status.
- #anthropic
- #claude
- #outage
- #api-reliability
- #logging
Tommy Rush — Founder, Rush Commerce
Operator turned builder. 15+ years running operations — now shipping the systems businesses run on. More
Get The Rush Report weekly — one email, zero fluff.
Keep reading
Trintech's finance agents execute inside your controls
Trintech launched three AI agents for financial close that do the work instead of recommending it — inside existing approvals and audit trails. Copy the pattern.
Read itSpaceXAI Team Bots: one shared agent holds the credentials
SpaceXAI opened Team Bots in public beta — shared AI agents with team memory, plugins, third-party credentials, and their own Slack handle. Scope them now.
Read it