Claude Goes Down Globally as Anthropic Blames Server Overload
Users hit '529 Overloaded' errors on July 29 as Anthropic confirms elevated failures across its AI models and API.
Key points
- Anthropic confirmed a global outage of its Claude AI assistant starting 7:49 p.m. UTC on July 29, 2025.
- Users saw an 'API Error: 529 Overloaded' message, meaning Claude's servers could not handle the volume of requests.
- Anthropic said at 8:33 p.m. UTC that it had identified the cause but did not share a fix time.
- The failure hit both the Claude chat product and third-party apps that call Claude through its API, the software bridge that lets other programs talk to Claude.
Anthropic's Claude chatbot went offline for users around the world on Tuesday evening, with the company confirming raised error rates across several of its AI models.
The outage started at 7:49 p.m. UTC on July 29, according to Anthropic's own status updates. Requests failed with a blunt message: 'API Error: 529 Overloaded. This is a server-side issue, usually temporary, try again in a moment.'
The issue was first reported by BleepingComputer, whose staff hit the same error repeatedly during testing.
By 8:33 p.m. UTC, Anthropic said it had identified the root cause and was working on a fix. It did not say what caused the surge, and it did not give a timeline for full recovery.
What is a '529 Overloaded' error?
It is Claude's way of saying its servers are swamped. The code sits in the same family as the more familiar '503 Service Unavailable' message websites show when they cannot cope. In plain terms, too many requests are arriving at once, and the system is turning some away rather than crashing entirely.
For everyday users, that meant Claude either refused to answer or cut off mid-response. For businesses, the impact was wider.
Why does this matter beyond the chat window?
Because Claude does not just power its own app. A growing number of companies build their own products on top of Claude's API, the connection that lets outside software send questions to Claude and get answers back. Customer support bots, coding assistants, document summarisers, and internal search tools all sit on that pipe.
When the API returns 529s, every downstream product returns errors too. A small business using a Claude-powered helpdesk tool would have seen its assistant go silent with no obvious explanation.
Timeline of the outage
| Time (UTC) | Event |
|---|---|
| 7:49 p.m., July 29 | Anthropic begins investigating elevated errors |
| Shortly after | Users report '529 Overloaded' failures worldwide |
| 8:33 p.m., July 29 | Anthropic says root cause identified, fix in progress |
| Not disclosed | Full recovery time |
Should users be worried about their data?
No, not based on what Anthropic has said so far. A 529 error is a capacity problem, not a security breach. There is no indication that anyone broke into Claude, stole conversations, or tampered with the models. The company has framed this as an availability issue, meaning the service could not be reached, rather than a confidentiality one.
Still, outages at AI providers are worth watching. More companies are wiring critical workflows into a handful of large model providers. When one of those providers stumbles, the ripple hits far beyond the provider's own front door.
Common questions
Is Claude working again now?
Anthropic said it identified the cause within about 45 minutes and was applying a fix. Users should check Anthropic's status page for the latest before assuming the service has fully recovered.
Do I need to change my password?
No. This was a capacity outage, not a hack. There is no security action for regular users to take.
What can businesses relying on the Claude API do during outages like this?
Build in a fallback. Route traffic to a second model provider, cache common answers, or show users a graceful holding message rather than a raw error.
