Claude Outage: Unable to Authenticate Errors Hit Thousands

Anthropic's Claude suffered a global authentication outage Sept. 29, drawing over 18,000 Downdetector reports before recovery by 10:59 a.m. ET.

Claude AI went down for many users on Tuesday morning, September 29, with complaints surging on Downdetector just after 10 a.m. ET. Users encountered error messages including "Claude unable to authenticate" and "temporarily unable to authenticate," with many unable to log in or start new conversations. The disruption affected claude.ai on desktop and mobile, Claude Code, Claude Cowork, platform.claude.com, and the Claude API. Reports climbed rapidly — Moneycontrol recorded 18,335 outage reports at 7:49 p.m. IST against a baseline of 10, while The Economic Times cited counts of more than 11,000 U.S. users.

While Claude has faced intermittent outages before, Goosed notes that authentication-specific failures at this scale are rare — most prior disruptions involved elevated model errors during high traffic rather than users being locked out of their accounts entirely. Some users reported intermittent access with frequent disconnections, others a full lockout. The scope also matters: an API outage affects third-party products built on Claude, and Claude Code and Cowork are increasingly used for development and business work. Notably, The Economic Times reported Claude for Government was the exception, appearing unaffected.

The outage trended on Google and spread across multiple regions. Goosed reported affected users in the United States, Europe, the Middle East and beyond. Downdetector grouped complaints with chat on claude.ai at roughly 45 percent of reports, the app at 29 percent, and Claude Code at 19 percent, according to International Business Times Australia, which also noted a gap between crowd reports and official status: at 10:19 a.m. ET Anthropic's status page had not yet listed a matching incident.

Anthropic acknowledged the incident on its status page, saying it was investigating elevated error rates across claude.ai, platform.claude.com, the Claude API, Claude Code and Claude Cowork, and that retrying a request might succeed. Moneycontrol reported the company later applied a mitigation, after which error rates dropped substantially, though some sign-ins, account actions, and Code and Cowork sessions were still failing. Mashable reported that per Anthropic's status page, all services were operating normally as of 10:59 a.m. ET, with The Economic Times noting most services had recovered as of 14:59 UTC.

The incident came less than a week after a prior disruption. The Next Web reported that on September 22 Anthropic saw a rise in errors affecting several models, including Opus 5, Mythos 5.1 and Fable 5.1, an incident resolved in about one hour and thirty-eight minutes per a third-party archive of status data. International Business Times Australia counted 79 incidents over 90 days across six components via Statusfield, describing a September pattern of short spikes rather than multi-day blackouts. A possible data-loss caveat: The Economic Times reported some messages sent between 14:00 and 14:59 UTC may not have been saved.

Anthropic has not stated the root cause of the disruption, though it said it would continue posting updates to its status page. For users and teams depending on Claude, the practical guidance from the pack is modest: check status.claude.com for official updates, avoid repeatedly refreshing to prevent compounding server load, and retry requests, since Anthropic itself noted retries may succeed. International Business Times Australia adds that teams building on the API should watch the official status feed alongside their own error dashboards, treating crowd reports as a signal to investigate rather than a verdict. With services restored, attention shifts to whether Anthropic explains what caused the authentication failures.