Anthropic Confirms 42-Minute Outage on Claude's Web Apps
TL;DR
- Anthropic's status page confirms authentication failures at 21:58 UTC on August 16 and full restoration by 22:40 UTC, roughly 42 minutes.
- Claude.ai, Claude Code, Claude Cowork, and platform.claude.com went dark while the Claude Console and Claude API stayed operational.
- The company has not disclosed a root cause, and BleepingComputer reports both incidents remained under investigation at press time.
A 42-minute window on August 16 knocked most of Anthropic's user-facing Claude surface offline while the API kept running, BleepingComputer reported. The status page trail is thin but specific: authentication started failing at 21:58 UTC, a broader service disruption with degraded performance hit Claude.ai and platform.claude.com shortly after, and services were restored by 22:40 UTC. Claude.ai, Claude Code, and Claude Cowork were all affected. The Claude Console and the Claude API held up.
The split matters more than the duration. If your team logs into Claude.ai to drive agent workflows or leans on Claude Code for development, the outage was total for the window. If you route through the raw API, you likely never noticed. That is a familiar failure mode for consumer-facing web apps that share an auth path, and it lands in a week where session-check tooling for Claude and its peers has been visible in AI Weekly's coverage the day before.
Anthropic told users it was 'investigating an issue preventing some users from authenticating,' then acknowledged a broader service disruption on the main properties. Whatever the trigger, the company has not disclosed it, and reporter Mayank Parmar writes that both incidents remained under investigation at press time. The article does not name an upstream identity provider, a configuration change, or a third-party dependency.
Two things the piece leaves open. The scale of impact, meaning how many paying customers or concurrent sessions were affected, is not quantified anywhere in the report. And 'restored by 22:40' does not describe whether users saw a clean recovery or a slow drain of queued requests, which matters if your CI or agent runs sat inside the failure window.
For teams running production Claude workloads, the practical read is boring but useful: architect around the API surface where you can, and treat Claude.ai and Claude Code as productivity layers rather than infrastructure. Expect more of these status-page moments as Anthropic's user base keeps expanding at the pace our tracker has recorded, which now spans 341 Anthropic stories in the last 90 days.
Originally reported by bleepingcomputer.com
Read the original article →Original headline: Anthropic Confirms 42-Minute Claude Outage Hits Claude.ai, Claude Code and Claude Cowork