ChatGPT, Claude, Grok and Cursor Hit by Widespread Outages

ChatGPT, Claude, Grok and Cursor Hit by Widespread Outages

ChatGPT, Claude, Grok and Cursor users reported widespread errors and failed requests on Thursday, September 3, as official status pages confirmed several AI services were experiencing outages or degraded performance.

ChatGPT, Claude and Grok outage reports
Users reported problems accessing ChatGPT, Claude and Grok amid simultaneous service disruptions on September 3, 2026. Image: X

The disruption affected consumer chatbots, artificial intelligence models, coding assistants and automated agents. However, there is currently no official evidence that the incidents were caused by a single technical failure or coordinated cyberattack.

Updated at 3:30 p.m. UTC on September 3, 2026: Some ChatGPT services were recovering, while incidents affecting Claude, Grok and parts of Cursor remained under investigation or unresolved.

ChatGPT Claude Grok and Cursor service outage status
Status information showing service disruptions affecting ChatGPT, Claude, Grok and Cursor on September 3, 2026. Image: X

ChatGPT, Claude and Grok outages confirmed

OpenAI said it began investigating elevated errors across ChatGPT and Codex at 2:58 p.m. UTC. In an update posted at 3:22 p.m. UTC, the company said it had applied a mitigation and was monitoring the recovery.

Users report major AI services outage
Social media users reported errors and access problems as several major AI services experienced disruptions on September 3, 2026. Image: X

The incident affected multiple ChatGPT and Codex components, according to the official OpenAI status page. OpenAI notes that availability can vary by subscription tier, model and API feature, meaning some users may regain access before others.

Anthropic separately confirmed elevated error rates across several Claude models. Its service-status dashboard listed Claude.ai, the Claude API, Claude Code and Claude Cowork as experiencing partial outages.

Anthropic said the incident began at about 1:26 p.m. UTC. The company later reported that most affected models had recovered to their normal error rates, but Claude Opus 4.8 and Opus 5 were still affected as of its 3:25 p.m. UTC update.

Grok was also experiencing a confirmed model outage. The official xAI incident page said the company began investigating at 1:30 p.m. UTC and was working to restore the service as quickly as possible.

Users took to social media to report blank responses, connection errors, failed prompts and unusually slow performance. The reports helped draw attention to the disruption, but the official company dashboards provide the most reliable confirmation of which systems were affected.

Why Cursor was also affected

Cursor reported two active problems. One involved a broader service degradation affecting Grok models, Automations, Cloud Agents, Grok Bot and Review Agents. A second incident involved elevated errors for Anthropic models used through the coding platform.

Cursor said the upstream Anthropic problem could cause failed requests or interrupted agent turns for users selecting affected Claude models. Its official status page listed degraded performance for the IDE, CLI, Cloud Agents and Automations, while Review Agents and Grok Bot were marked as suffering major outages.

This overlap helps explain why several AI tools appeared to fail at roughly the same time. Cursor relies on models and infrastructure supplied by outside AI companies, so problems at Anthropic or xAI can also disrupt features inside Cursor. That does not prove the OpenAI, Anthropic and xAI incidents share the same root cause.

The latest disruption follows a series of high-profile online service problems affecting work and communication tools. Swikblog recently reported on a separate Microsoft Outlook outage that disrupted users, highlighting how quickly interruptions can affect people who rely on cloud platforms for daily work.

Users experiencing problems should avoid repeatedly submitting the same prompt or API request, particularly when usage-based billing applies. Saving unfinished work locally and checking the provider’s official status page before troubleshooting devices, browsers or internet connections may also prevent unnecessary changes.

None of the affected companies had announced a security breach or provided evidence of a malicious attack at the time of publication. The incidents should therefore be described as concurrent service outages unless the companies later identify a common cause.

This is a developing story. Service conditions may change quickly as OpenAI, Anthropic, xAI and Cursor deploy fixes and publish further updates.

Add Swikblog as a preferred source on Google

Make Swikblog your go-to source on Google for reliable updates, smart insights, and daily trends.

Get the Swikblog App
Stay updated with breaking news, trending stories and the latest updates—all in one place.
Get the Swikblog app on Google Play
Free download for Android devices