Why Anthropic Claude is down today and how to check service status 

The "Claude will return soon" hard block screen users are hitting when attempting to log in. Image: Tech Bird

Anthropic’s Claude AI is experiencing a major global outage today, March 2, 2026. A famous AI, especially built for developers, is now facing a major login failure and outage disruption all around the world. If you are staring at a completely unresponsive chat interface or getting booted to a stark "Claude will return soon" error page, trust me, you are not alone.

A severe service disruption is actively knocking claude.ai, the developer console, and Claude Code offline for thousands of users worldwide. People attempting to access their workspaces are just being met with widespread login failures and persistent error banners.

Why this matters right now: 

This isn't just a slow server. It means active, mid-project workflows are dead in the water for anyone relying on the web interface.

Just this evening, I was about to try running a prompt using the Sonnet 4.6 Extended model, only to be met with a yellow warning banner stating, "This isn't working right now. You can try again later." before eventually hitting a hard logout.

Here is exactly what we know about the Claude outage today, why the interface looks the way it does right now, and what parts of the platform are actually still functioning.

What is causing the Claude outage today?

Anthropic is officially tracking the incident. They initially flagged "elevated errors" across all of their platforms, but the exact nature of this Anthropic server status is becoming much clearer right now.

The core issue is not a failure of the underlying Large Language Models (like Sonnet 4.6 Extended or Opus 4.6). Instead, the disruption is rooted directly in Claude's identity and authentication layer. Anthropic engineers are isolating the problem specifically to the login and logout paths.

This is why the Claude AI down reports are showing different impacts for different users:

• Claude.ai Web Interface: Close to 75% of Downdetector complaints are directly related to the chat service. The web interface relies heavily on active session management, which is currently failing.
• Claude Code and Mobile App: Mobile app users make up 13% of the reported issues, while Claude Code accounts for 12%.
• Claude API: The official status page confirms that the Claude API is working exactly as intended. Automated systems and third-party apps relying on Claude's backend via static API keys are largely unaffected because they bypass the web-based OAuth login flow entirely.

Decoding the specific errors you are seeing

If you are trying to access your workspace right now, you are likely encountering one of two distinct screens based on your current session status:

The "This isn't working right now" banner

Your workspace might still load visually, but prompts are being blocked by a yellow warning. Image: Tech Bird

If you were already logged in, your interface might still load the basic chat UI ("Good evening, Zeeshan Saleem") and your model selector. However, any attempt to actually send a prompt is immediately blocked by a yellow warning banner at the top right corner stating: "This isn't working right now. You can try again later." This happens because your local browser session is visually active, but it cannot securely communicate with Anthropic's backend servers to process the prompt.

The "Claude will return soon" hard block

If you attempt to refresh your broken session, log out, or log back in, the system completely drops you. You will hit a dark grey screen with the Anthropic logo stating: "Claude is currently experiencing a temporary service disruption. We're working on it, please check back soon." The provided "Try again" button will simply loop this page until the authentication paths are fully restored.

The ChatGPT comparison and Claude's massive user surge

Anthropic is actively investigating the root cause and working to implement a fix. While they haven't provided a strict ETA for when full web service will be restored, server overload issues of this scale are often a byproduct of surging user demand.

Here is the real comparison: 

When OpenAI's ChatGPT faced massive user influxes during its peak, their authentication layers regularly crumbled exactly like this. This technical failure arrives during a week of massive, unprecedented growth for Anthropic. Claude recently dethroned its main competitor, OpenAI's ChatGPT, in Apple's App Store, hitting the number one spot as the most-downloaded free app in the U.S. This massive influx of new users adopting the platform has naturally put immense strain on Anthropic's infrastructure. In total, the outage-tracking website Downdetector logged close to 2,000 user reports during the initial crash alone.

My Verdict: What to do next

If your workflow is completely stopped, the most effective immediate workaround is to utilize the Claude API if you have existing keys configured, as it completely bypasses the broken web authentication. Otherwise, my advice is to hold off on constantly clicking "Try again" on the error page. You are only contributing to the server overload. Wait until the official status page confirms the fix is live. Skip the web interface for tonight, pivot to your API if you have deadlines, or just take a break.

Update (14:42 UTC): Issue Resolved

Anthropic has officially resolved today's service disruption. At 14:35 UTC, engineers began investigating a brief secondary wave of "elevated errors" specifically impacting the Claude Opus 4.6 model. Fortunately, by 14:42 UTC, the root cause was identified and a permanent fix was implemented across the authentication layer. The login paths and web interface are now fully operational again, so you can safely refresh your workspace and get back to prompting!