ChatGPT
ChatGPT

Several of the world's most widely used artificial intelligence services experienced simultaneous outages Thursday, with ChatGPT, Claude, Grok, Gemini and the AI coding tool Cursor all logging spikes in user-reported problems within roughly the same half-hour window, according to outage-tracking site Downdetector.

Downdetector said Thursday's disruption to OpenAI's ChatGPT drew more than 340,000 reports globally, marking the platform's largest outage by Downdetector report volume in more than a year. The scale of the disruption, and the unusual coincidence of multiple competing AI platforms going down at once, drew significant attention from users and industry observers, given how unusual it is for services built on distinct backend infrastructure and operated by rival companies to fail within the same narrow time window.

According to OpenAI's status page, the ChatGPT outage began around 10:58 a.m. ET Thursday, ultimately affecting 15 separate ChatGPT components, including login, search, image generation, voice mode and its Deep Research feature, along with four components of OpenAI's Codex coding tool.

Anthropic's Claude experienced a parallel disruption around the same time. According to the company's status dashboard, elevated errors affected a specific set of Claude models, including Opus 5, Opus 4.8 and Opus 4.6, spanning the Claude.ai web interface, Anthropic's developer API, and its Claude Code and Claude Cowork products. Anthropic later reported that most Claude models had recovered to baseline error rates even as some of the more advanced models continued experiencing issues for a longer stretch.

Grok, the chatbot developed by Elon Musk's xAI, was also affected, with the company's own status page confirming an active outage impacting both its web application and API across two separate regions. According to Bloomberg, xAI acknowledged the issue and said it was working to address the problem. SpaceX later attributed the underlying cause of the Grok disruption to an outage at the company's computing center in Memphis, Tennessee.

Cursor, a popular AI-powered coding tool that relies on underlying models from multiple providers, was also swept into the disruption, with the company citing the concurrent Grok and Claude issues as the source of its own service problems.

Google's Gemini experienced a comparatively narrower disruption than the other affected platforms. While Downdetector recorded a spike in user reports related to Gemini, and some users reported connection timeouts and internal server errors, Google never issued an official confirmation of a full platform outage. Google AI Studio's status page did note problems specifically affecting the serving of newly created API keys, which the company described as a partial degradation rather than a complete service failure.

Recovery timelines varied somewhat across the affected platforms. According to tracking from 9to5Google, services began recovering around 8:49 a.m. Pacific time, with full restoration confirmed across ChatGPT, Claude and Grok by 12:38 p.m. Pacific time on Sept. 3. Separately, tracking site EWN reported that most affected systems had returned to normal performance by roughly 7 p.m. South African time, with OpenAI describing the underlying issue behind the ChatGPT and Codex disruption as a routing error that the company said it fixed within about an hour.

Despite the scale and unusual synchronicity of Thursday's outages, none of the companies involved, including OpenAI, Anthropic or xAI, published a detailed public root-cause explanation identifying precisely why so many separate AI platforms experienced disruptions within such a tight window. No infrastructure vendor, distributed denial-of-service attack, or specific shared configuration bug was publicly identified as a common underlying cause connecting the various outages.

The disruption prompted widespread commentary on social media, with some users noting the rare experience of being temporarily unable to rely on AI tools for everyday tasks. One social media user, writing on Bluesky, offered a wry observation about the moment.

"For its part, OpenAI's status page listed 15 affected ChatGPT components," one report noted, capturing the scale of the disruption across the platform's various features, while a widely shared social media post remarked that "for a brief moment, millions of people had to use their brains again."

Industry commentators have pointed to Thursday's outages as a notable moment for a technology sector increasingly dependent on cloud-hosted AI services for both consumer and enterprise workflows. One analysis noted that the disruptions affected multiple distinct backend architectures, regional data centers and specialized programming interfaces simultaneously, a scale of impact that exceeded what would typically be expected from a routine single-vendor cloud hiccup.

Some technology commentators used the incident to argue for greater resilience planning among businesses that have built critical workflows around cloud-based AI tools. Suggestions circulating in the aftermath of the outage included accelerating the deployment of smaller, locally hosted open-source AI models as fallback options during cloud outages, and implementing more aggressive timeout thresholds and automated fallback mechanisms within software systems that depend on external AI application programming interfaces, to prevent operations from stalling entirely when a given provider experiences a disruption.

Thursday's incident adds to a growing list of significant AI service disruptions recorded throughout 2026, a year that has already seen several major single-vendor outages affecting individual AI platforms, including a Claude-specific incident in June that lasted approximately seven hours. The unusual coincidence of multiple major providers experiencing outages within the same narrow window, however, distinguishes Thursday's disruption from those earlier, more isolated incidents.

For users and businesses that experienced disruptions Thursday, the affected companies' official status pages, including status.openai.com for ChatGPT and Codex, along with Anthropic's and xAI's respective status dashboards, remain the most authoritative sources for confirming whether any new related incidents emerge going forward. As of the latest available updates, all major affected platforms had returned to normal operating status, though none of the companies involved had provided users with a comprehensive explanation of what caused so many independent AI systems to falter within the same brief window on Thursday morning.

Read more