Three of the world's most widely used AI platforms—OpenAI's ChatGPT, Anthropic's Claude, and SpaceX's Grok—went dark simultaneously on Thursday in a rare multi-hour outage that left enterprises scrambling and exposed a critical vulnerability in modern workflows. According to a report published Thursday by Computerworld, the incident reveals how quickly companies have automated work with generative AI agents without preparing for inevitable failures. The outages highlight an uncomfortable truth: businesses are deploying AI faster than they're building backup plans.
The disruptions began Thursday morning Eastern time and lasted hours. ChatGPT services experienced problems starting around 11 a.m. ET and remained down for roughly two hours, impacting search functions, file uploads, agents, GPTs, voice mode, image creation, ChatGPT work, Compliance API, Deep Research, ChatGPT Atlas, and various connectors and applications, according to the report. Users couldn't log in, conversations wouldn't load, and the interface threw errors when people tried to send messages. OpenAI's Codex services—including web, API, command line interface, and VS code extension—also failed. OpenAI restored functionality by 12:55 p.m. ET. Claude began failing around 7:37 a.m. ET and stayed offline for approximately four hours, with Anthropic acknowledging elevated errors across multiple models: Mythos and Fable 5.1 and 5, Sonnet 5, and Opus 5, 4.8, and 4.6. The platform recovered by 11:27 a.m. ET, just one day after a separate 27-minute outage. Grok hit problems around 9:30 a.m. ET, affecting Grok Web, Build, API, Office/Workspace plugins, Android, and X, and didn't return to healthy traffic until 1:08 p.m. ET—nearly three and a half hours later.
Technology analyst and journalist Carmi Levy told Computerworld that the incident should "serve as a wakeup call to IT leaders who have largely ignored what it'll cost them if these increasingly critical platforms suddenly go dark." The report notes that AI agents are now taking over automated and wider-scale workflows, leaving enterprises potentially "uncomfortably exposed" when systems fail. Brian Jackson, a principal research director at Info-Tech Research Group, called the simultaneous outages "a curious scenario," theorizing the cause might be related to shared infrastructure like a content delivery network layer, domain name system, or common cloud infrastructure.
The synchronized failures matter because the role of AI in business has fundamentally shifted in recent months. Just a short time ago, most employees used AI chatbots only for basic questions or drafting simple emails, so outages had minimal impact on organizational productivity, the report explains. But as companies hand more complex tasks to cloud-based automated tools, the stakes have climbed sharply. When Thursday's outages hit, workers had to fall back on traditional manual methods like updating spreadsheets or assembling reports the old-fashioned way—only to discover that after relying on AI agents to handle so much work, their cognitive skills might not be as sharp as they once were, Levy noted. Unlike cloud productivity platforms such as Google Workspace and Microsoft 365, which offer limited offline modes using locally stored data, agentic AI platforms provide far fewer offline workarounds in their current form.
The report recommends that organizations revisit disaster recovery and business continuity plans to assess the productivity impact of potential service outages. Companies should document workflows in greater detail and scenario-plan what near-term recovery would look like during an extended AI platform failure. Better training is essential to ensure employees maintain their manual skills over time and can deploy them during service interruptions, because the more businesses lean on agents to complete critical tasks and remove humans from the loop for efficiency's sake, the less capable workers will be to step back in during inevitable disruptions. Jackson advises enterprises to adopt a modular architecture for large language models, viewing the model as "a commodity that can be hot-swapped with an alternative"—whether that's another cloud provider or a self-hosted option like an open-weights model. Levy warns that "too many organizations are about to learn some hard lessons about not having a backup plan in place." Companies that treat AI resilience as optional rather than essential may discover that operational continuity now depends on technology layers they never stress-tested. The question isn't whether these platforms will fail again, but whether businesses will be ready when they do.