Claude Opus 4.6 outage on March 17: what users need to know

⚠️ As of Mar 17, 2026 at 19:47 UTC, Anthropic's public status page lists the Claude Opus 4.6 incident as identified, not resolved.

Claude Opus 4.6 is down for a slice of users right now, and this one is showing up in exactly the places teams care about: the chat app, the developer platform, the API, and Claude Code.

Anthropic's public incident feed says the issue was posted at 19:47 UTC on March 17, 2026 with this update: "The issue has been identified and a fix is being implemented." The affected components listed by Anthropic are claude.ai, platform.claude.com, the Claude API, and Claude Code.

What users are seeing

The first sign for many users is not a subtle slowdown. It is a hard failure — timed 500 internal server errors coming back from the Claude API.

That matters because a 500 is not just an annoying retry in the chat box. It can break app flows, scheduled jobs, coding sessions, and any tool chain built on the Anthropic API.

What the official Claude site shows

Anthropic's own status site confirms the incident is live and currently in the identified phase. The status page says the issue has been identified and a fix is being implemented. At the moment, Anthropic is not describing this as a narrow issue affecting only one surface. The degraded components listed are the web app, the developer platform, the API, and Claude Code.

Why this outage matters

When an AI model outage hits both the product and the API at the same time, the blast radius is much larger than a normal chatbot hiccup. Consumers lose access in the app, developers see requests fail, and teams using Claude in production workflows need retries or fallback logic immediately.

The bigger story is reliability. Model quality gets the headlines, but uptime is what decides whether people can actually ship on top of a platform. Incidents like this are why serious teams treat model redundancy, backoff, and graceful degradation as core product features.

Current facts

Status: Identified
First public timestamp: Mar 17, 2026 at 19:47 UTC
Anthropic update: "The issue has been identified and a fix is being implemented."
Affected surfaces: claude.ai, platform.claude.com, the Claude API, and Claude Code

What users should do now

Expect intermittent failures or hard API errors until Anthropic closes the incident.
Re-run failed requests rather than assuming the first error is permanent.
If Claude powers customer-facing workflows, switch to retries, queueing, or alternate model fallbacks now.
Keep the Claude status page open if you are actively using Opus 4.6 today.