Claude AI Down Today: Why Anthropic's AI Stopped Working and What Happened to User Quotas
If you tried opening Claude today and ran into failed requests, endless loading screens, or a mysteriously depleted usage quota — you were not alone. On June 2, 2026, Anthropic's Claude AI suffered a significant service disruption that hit thousands of users across the globe, affecting everything from the web app to the API and developer tools.
Here is a full breakdown of what happened, why it happened, and what Anthropic did about it.
What Happened: Claude AI Goes Down on June 2, 2026
The first signs of trouble began surfacing around 2:10–2:19 AM Eastern Time, when a sharp spike in user complaints appeared on Downdetector. Within minutes, reports climbed toward 3,000 — a clear signal that this was not an isolated glitch.
The disruption was widespread. Affected services included:
Claude.ai (web platform)
Claude Mobile App (iOS and Android)
Claude API (used by developers and businesses)
Claude Code (the AI coding assistant)
Claude Console
Anthropic officially acknowledged the problem through its status page at 6:04 UTC, initially posting "Investigating — We are currently investigating this issue." Updates followed at 9:33 UTC ("Identified — a fix is being implemented") and the incident was finally marked Resolved at 11:49 UTC.
The primary model flagged in Anthropic's status updates was Claude Opus 4.6, though reports suggest the disruption was not limited to a single model.
Why Did Claude AI Go Down? The Sub-Agent Bug Explained
This was not a routine server crash. The root cause, according to multiple reports, traces back to a bug inside Claude Code's sub-agent system.
What Are Sub-Agents?
Claude Code handles complex developer tasks by splitting them into smaller parallel operations managed by "sub-agents" — essentially mini AI processes that work simultaneously to speed up results. It is an elegant system under normal circumstances, allowing Claude to tackle large codebases and multi-step tasks far more efficiently than a single sequential process would allow.
But when something goes wrong with how sub-agents are created, things unravel fast.
The Bug: Runaway Sub-Agents
In today's incident, a malfunction caused Claude Code's sub-agents to multiply uncontrollably. Instead of spawning a limited number of helper processes per task, the system kept generating more and more — far beyond what any task required. Each one consumed tokens in the background, and the effect compounded quickly.
Think of it this way: if one task was supposed to use five sub-agents, the bug was creating fifty or five hundred. All of them burning through resources simultaneously, whether the user noticed it or not.
The "Silent Drain" — Why Quotas Disappeared So Fast
For many users, particularly those on Pro and Max plans, the most alarming part of the outage was not the downtime itself — it was watching their usage quota evaporate in minutes.
This became known informally as a "silent drain." Unlike a complete service blackout, the platform was still technically running. Users were prompting Claude, getting responses, or running Claude Code tasks — but behind the scenes, the runaway sub-agents were consuming tokens at a rate that had nothing to do with actual usage.
Some Max plan subscribers reported that quotas designed to last several hours were exhausted within just a few minutes of activity. Developers working on large projects were hit especially hard, finding themselves locked out mid-task with no quota remaining and no clear explanation at the time.
According to Downdetector data, the majority of user complaints centred on Claude Chat (60%), followed by the mobile app (24%), and Claude Code (8%).
How Anthropic Responded
Once the volume of complaints became clear, Anthropic moved on two fronts simultaneously.
Fixing the Bug The engineering team identified the sub-agent malfunction and began rolling out a patch. Status page updates reflected steady progress through the morning UTC hours, and the fix was confirmed deployed by late morning.
Emergency Quota Reset Recognising that affected users had lost usage credits through no fault of their own, Anthropic initiated an automated emergency quota reset for impacted Pro and Max plan accounts. This restored depleted tokens back to user balances — essentially compensating subscribers for what the bug consumed rather than what they actually used.
Anthropic has stated it has not yet released a full technical post-mortem, but confirmed that the quota reset was specifically tied to the bug and was not a general promotional credit.
Who Was Most Affected?
Developers using Claude Code bore the brunt of today's disruption. Those in the middle of long coding sessions, debugging workflows, or multi-file refactoring tasks found their projects grinding to a halt the moment their quota hit zero. Some had no idea the quota was draining until they were already locked out.
Businesses and teams that have built Claude into daily workflows — content pipelines, customer support tools, internal automation — also experienced unexpected productivity gaps, even if just for a few hours.
Students and individual users on free or standard plans noticed slower responses and intermittent failures, though the quota drain issue was primarily felt by paid plan holders.
Is Claude AI Back Up and Running?
Yes. As of the official status update at 11:49 UTC on June 2, 2026, the incident has been marked resolved. Anthropic says it is continuing to monitor system performance to ensure stability holds.
That said, some users may notice slightly inconsistent response times as the fix fully propagates across Anthropic's infrastructure. If you are still experiencing issues, the best place to check is status.claude.com for real-time service status.
If your quota was affected and you have not yet seen it restored, it is worth checking your account dashboard or reaching out to Anthropic support directly.
A Broader Lesson: The Risks of AI Dependency
Today's outage sparked a wider conversation about what happens when AI tools we rely on for critical work suddenly go dark. As TechRadar put it, the incident is a reminder that promised productivity gains from AI still run through a single point of failure: someone else's status page.
With more companies, development teams, and individual professionals embedding Claude into daily workflows, the cost of downtime is no longer just inconvenience — it translates directly into missed deadlines, stalled deployments, and lost revenue.
Building in fallback options — whether that is an alternative AI provider, offline tooling, or simply not scheduling critical deliverables around AI-only workflows — is becoming less optional and more necessary.
Quick Recap: What You Need to Know
Claude AI went down on June 2, 2026, starting around 2:19 AM ET
The cause was a bug in Claude Code's sub-agent system that generated runaway processes and consumed tokens at an abnormal rate
Pro and Max plan users experienced rapid and unexpected quota depletion — sometimes losing hours of allowance in minutes
Anthropic issued an emergency quota reset for affected paid subscribers
The incident was officially resolved at 11:49 UTC
Check status.claude.com for real-time updates if you are still seeing issues
Comments (0)
No posts yet
Be the first to comment!