Claude partial – Claude users across web and mobile reported problems beginning around 2.10am ET / 7.10am GMT, with a spike showing up on Downdetector. Anthropic confirmed a partial outage, citing elevated errors on Opus 4.6 and “unexpected capacity constraints,” before later
The morning starts like any other—until your chatbot starts stalling.
For many users, Claude stopped behaving as expected as early as around 2.10am ET / 7.10am GMT. Downdetector began showing a spike in reports, and at the time of writing the numbers were still climbing, reaching 216 reports in the UK. In the US, reports had also reached 139.
Anthropic, Claude’s developer, confirmed the issue was real and active work was underway. The Claude Status page said there was a “partial outage,” and that a “fix is being implemented.”
The problems showed up most clearly in Claude Chat. Other services—Claude Cowork and Claude for Government—seemed to be unaffected.
Over the next 90 minutes, the picture looked unsettled but not chaotic. Reports in the UK had been steadily climbing, with signs that they were dropping slightly. That shift seemed to line up with Anthropic’s message that it was still working on a fix.
But another update on the status page suggested this might not be a quick recovery. It reported “elevated errors on Opus 4.6,” Claude’s latest flagship model announced in February. Anthropic added that “we are continuing to work on a fix for this issue.”
By the time three hours had passed since the announcement that a fix was being implemented, it was still proving tricky.
Even if Opus 4.6 was explicitly flagged, the outage didn’t behave like a problem limited to one model. One user attempting to use Sonnet 4.6 said it hung indefinitely with messages including “gathering my thoughts” and “still working on it.” Others reported Claude being very slow that morning. even as Downdetector figures started to move in the right direction.
For some, the disruption carried an awkward extra message. When trying the free plan, Claude issued a warning that, “due to unexpected capacity constraints, Claude is unable to respond to your message.” It then suggested upgrading to a Pro plan.
That suggestion landed poorly with users because, if the outage was truly ongoing and categorized as a partial outage, it meant a paid upsell appeared right when free access was failing.
The outage’s reach also raised a bigger question for people relying on Claude as more than a single chat window. The reported disruption seemed to affect Claude Code as well as the chat interface on web and mobile.
A note shared by one user captured the frustration in plain terms: “Claude is down. It’s a nice reminder that the promised 10x productivity gains still have a single point of failure: someone else’s status page… 🫠— @P3b7_ (June 2, 2026)”
Earlier this year, more users had moved toward Claude after some dissatisfaction with competing tools. The story around the switch—some people moving from ChatGPT to Claude due to the former’s unpopular AI military deals—adds another layer to what today’s outage means: people don’t just rely on one assistant out of habit. they rely because they’ve decided it fits their workflow.
When the service failed, the impact was felt immediately.
Even as the Downdetector numbers stabilized—around 145 reports in the UK at one point—the latest from Anthropic kept pointing to continued work. The status page still described the situation as a “partial outage. ” and the repeated language was that it was “continuing to work on a fix for this issue.”.
There was a point when the UK graph appeared to be trending downward slightly earlier in the morning. but the counts soon leveled out again. hovering around 150 reports in the UK and about 100 in the US. The message stayed consistent: a fix was in progress, and the outcome would show up in the incoming reports.
That change did come.
A later update on the Claude status page said “a fix has been implemented and we are monitoring the results.” At least for some users, the service began to respond again. One report described Claude still not working on the web at first, but the direction of travel was promising.
By the time conditions improved, the pattern on Downdetector also looked like recovery. Reports dropped sharply—to around 30 in the UK—while the US numbers fell as well. Anthropic also updated the Claude Status page with an “all systems operational” message.
If there was a lesson in the outage, it didn’t come from a statement. It came from the timing. the stalled messages. the capacity warning for free users. and the simple fact that when one widely used AI assistant stumbles. it doesn’t affect only one conversation. It ripples across everything built on top of it.
Claude outage Anthropic Opus 4.6 Claude Status page Downdetector Claude Chat unexpected capacity constraints