STATUS — RESOLVED
The June 23 outage is now resolved. All Claude platforms are operational. Check status.anthropic.com for real-time status and status.claude.com for component-level detail.
OUTAGE FACTS — JUNE 23, 2026
● Started: 14:19 UTC (10:19 AM ET / 2:19 PM BST / 9:19 PM ICT)
● Identified: 14:25 UTC — 6 minutes after start
● Resolved: Confirmed resolved — see status.anthropic.com
● Affected: claude.ai · Claude Console · Claude API · Claude Code · Claude Cowork
● NOT affected: Claude for Government
● Severity: Elevated error rates across all models simultaneously
● User reports: 1,272 submitted to StatusGator in 24 hours
● Root cause: Not disclosed publicly by Anthropic
What Happened — Timeline
| Time (UTC) |
Time (ET) |
Status update |
| 14:19 UTC |
10:19 AM ET |
Anthropic begins investigating — "elevated error rates across multiple models" |
| 14:25 UTC |
10:25 AM ET |
Identified — fix being implemented. All 5 platforms confirmed affected |
| ~15:25 UTC |
~11:25 AM ET |
Still unresolved per TechRadar live blog — fix not fully working, "continuing to monitor" |
| Resolved |
— |
Outage confirmed resolved — all platforms operational |
What Was Affected
DOWN: claude.ai (web + mobile), Claude Console (platform.claude.com), Claude API (api.anthropic.com), Claude Code (terminal), Claude Cowork
UNAFFECTED: Claude for Government — the only platform that remained fully operational throughout the outage
The outage hit all Claude models simultaneously — not just one model tier. Users on Sonnet 4.6, Opus 4.8, Haiku 4.5, and the recently restored Fable 5 all reported errors. The error pattern: requests either failed immediately with 529 Overloaded errors or appeared to process (showing thinking indicators) but returned no response and ultimately timed out.
Context — Third Major Incident in June
This is the third significant Claude service disruption in June 2026, and the pattern is drawing attention. June has already turned into a test of how much instability users will forgive.
| Date |
Incident |
Duration |
Scope |
| June 12-18 |
Fable 5 + Mythos 5 suspended — export control directive |
6 days |
Fable 5 + Mythos 5 globally — government-mandated |
| June 22 |
Elevated errors for Opus 4.8 (multiple separate incidents) |
1h 50m + 1h 30m |
Opus 4.8 primarily, some Sonnet 4.6 |
| June 23 (today) |
Elevated errors across ALL models — all platforms |
60+ minutes |
All models · claude.ai · API · Code · Cowork |
The June 23 outage is bigger than June 22 in scope — June 22 primarily hit Opus 4.8, while today's incident hit all models and all platforms simultaneously. The previous last significant all-platform outage was June 2. For engineers and enterprises running Claude-dependent production workloads, three weeks between major outages — with a 6-day model suspension in between — is a reliability pattern that demands attention.
What to Do During a Claude Outage
Check status.anthropic.com first. Anthropic's official status page updates within minutes of incidents. The component breakdown shows which specific services (API, claude.ai, Claude Code) are affected — useful for knowing if your specific workflow is impacted.
Immediate fallbacks by use case: For API work — OpenRouter automatically routes to working providers (Gemini, GPT-5.5) with no code changes. For coding — Cursor (still available, routes to GPT-5.5) or Windsurf. For chat — ChatGPT or Grok on grok.com.
The 529 Overloaded error means the server is rejecting requests due to capacity — not a credentials or billing issue. Retry after 30-60 seconds. If retries consistently fail for more than 5 minutes, treat as an outage and use fallbacks rather than hammering the API.
For production systems: Bifrost (open-source AI gateway) automatically detects Anthropic degradation and routes to healthy alternative providers with zero code changes. The June pattern suggests this is worth implementing before the next incident rather than after.
The Bigger Picture — Reliability in June 2026
The outage pattern in June reflects two pressures converging on Anthropic simultaneously. The first is demand: Claude usage is growing faster than infrastructure can absorb it comfortably — the Fable 5 launch on June 9 drove massive spikes, the restoration on June 18 drove another, and ongoing enterprise adoption is steadily raising the baseline load. The second is complexity: running multiple model tiers (Haiku, Sonnet, Opus, Fable 5, Mythos 5) across multiple platforms (API, claude.ai, Claude Code, Cowork, Console) with nationality-based access controls added June 18 is significantly more operationally complex than the pre-June infrastructure.
Anthropic's response time is creditably fast — 6 minutes from investigating to identified today is excellent incident management. The open question is whether faster identification translates to shorter total outage duration as June's complexity increases. The September 2026 earnings call will be the first time Anthropic publicly addresses reliability metrics under public scrutiny.
Frequently Asked Questions
Is Claude down right now?
The June 23 outage is resolved. For current status, check status.anthropic.com — this is Anthropic's official status page, updated in near real-time. The component breakdown shows claude.ai, API, Claude Code, Console, and Cowork status individually.
Why is Claude showing a 529 error?
529 Overloaded is a server-side error meaning Anthropic's infrastructure is rejecting requests due to capacity. It is not a billing, credentials, or account issue. During an outage it affects all users equally regardless of plan. Wait 30-60 seconds and retry. If the error persists for more than 5 minutes, check status.anthropic.com and use a fallback provider.
What is the best Claude fallback during an outage?
For API work: OpenRouter with automatic failover to Gemini 2.5 Flash ($0.30/M) or GPT-5.5. For coding: Cursor or Windsurf — both route to GPT-5.5 and Gemini and were unaffected. For chat: ChatGPT (chatgpt.com) or Grok (grok.com). For production systems that cannot tolerate manual failover: Bifrost open-source AI gateway handles automatic multi-provider routing.
Why was Claude for Government unaffected?
Claude for Government runs on separate, isolated infrastructure from the commercial Claude platforms — a standard practice for government cloud services that require higher availability SLAs and regulatory compliance. This separation means government infrastructure incidents and commercial incidents do not cross-affect each other. The same isolation that protected Claude for Government from the June 23 outage is what kept Mythos 5 access available to cleared government users during the June 12-18 commercial suspension.
Sources: status.anthropic.com · status.claude.com · StatusGator · The Next Web · Related: Claude News 2026 hub · Best Claude Code alternatives · June 23 AI news roundup