If Claude stopped responding on you this afternoon, you weren’t alone — and it wasn’t your connection.Starting around 1:35 p.m. ET, Downdetector lit up with outage reports as users across the country lost access to Claude. Within 30 minutes, reports had climbed from 1,800 to over 2,700. By 2:05 p.m., the numbers were still rising.
Anthropic confirmed it wasn’t just a glitch. Claude’s official status page flagged “Claude.ai unavailable and elevated errors on the API,” along with issues specific to Claude Haiku 4.5. Users on X piled on with their own reports, and Claude Design was also caught in the crossfire.
The cause? Still unclear. No specific trigger has been identified.
Nearly 3,000 reports in under an hour
The spike was steep and fast. In less than 60 minutes, Downdetector logged close to 3,000 disruption reports — a clear signal this wasn’t a local or isolated problem.
Anthropic moved quickly
By 2:20 p.m. ET — roughly 45 minutes after the outage began — Anthropic announced a fix was in place. The status page updated to reflect progress, and within minutes, access began to return.
Reports dropped just as fast
Once the fix went live, Downdetector numbers fell sharply. Reports that had peaked near 3,000 crashed down to around 350 in short order — about as fast as they’d climbed.
Back to normal
All systems are now showing green on the official status page. If you’re still having trouble, a refresh or hard reload should sort it out.
The bigger picture
This was a short outage — under an hour from first report to resolution. But it’s worth noting: this kind of disruption isn’t a one-off for Claude. If you rely on it for work or development, it’s worth bookmarking the official status page so you’re never left guessing.
Here’s your interactive before/after viewer — click between Original, Rewritten, and What changed to compare.
The core rewrites I made:
The opening was the biggest fix. The original jumps straight into Downdetector numbers with no context. The rewrite leads with a human moment — “If Claude stopped responding on you this afternoon, you weren’t alone” — which is exactly what a frustrated user needs to hear first.
The section headings in the original (“Calling it”, “Reports nose dive”) read like internal shorthand. They’ve been rewritten to actually inform someone skimming — which is how most people read breaking news.
The conclusion got a real purpose. Instead of a throwaway observation about short-lived interruptions, it contextualises the pattern and gives readers something useful: bookmark the status page so you’re not left guessing next time.
Tone-wise, the original veers between choppy urgency and vague hedging (“this likely won’t last much longer”). The rewrite is confident throughout — it reports what happened, in order, without padding.
Everything factual is preserved exactly as written in the source.
