Anthropic’s popular AI assistant, Claude, experienced a significant disruption today, with users worldwide reporting that both web and code-related services were slow, unstable, or completely unavailable.
The incident began around 2:10am ET (7:10am GMT), when Downdetector showed a sharp spike in problem reports from multiple regions. At one point, reports climbed to over 200 in the UK and around 140 in the US, indicating a broad impact across free and paid users.
Claude Down for Users
Anthropic’s official status page initially listed the incident as a “partial outage,” confirming elevated error rates across key Claude services. The company stated that a fix was being implemented.
It later updated the page to note that a remediation had been deployed and was being actively monitored. Despite this, many users continued to see failed or delayed responses, suggesting that recovery was gradual rather than instantaneous.
The outage did not appear limited to a single interface, with users reporting issues on Claude’s web chat, mobile access, and Claude Code.
Anthropic later acknowledged “elevated errors on Opus 4.6,” the flagship model released in February, while users also reported that Sonnet 4.6 sessions hung indefinitely, with “gathering my thoughts” and “still working on it” messages. For some, the service would eventually respond after long delays; for others, requests failed outright or timed out.
One particularly controversial behavior emerged on the free tier, where some users received a warning that “due to unexpected capacity constraints, Claude is unable to respond to your message,” followed by a prompt to upgrade to a Pro plan. Many saw this as poorly timed, given that the root issue appeared to be platform instability rather than normal capacity limits.
Over time, Downdetector graphs began to show signs of improvement, with UK reports dropping from the hundreds to a few dozen, and US reports also declining as mitigation efforts took effect. Anthropic’s status page eventually moved to “all systems operational,” although some users still observed slower-than-normal responses and intermittent failures shortly after the update.
From a reliability and cybersecurity operations perspective, the incident highlights a growing risk: heavy dependence on a single AI provider for critical workflows.
As more organizations shift from one AI vendor to another, such as recent moves from ChatGPT to Claude, outages like this demonstrate the need for multi-vendor strategies, fallbacks, and clear incident communication when core AI infrastructure has a bad day.
