ChatGPT suffered two separate outages within a single 24-hour window, leaving users across the United States, the United Kingdom, and several other countries scrambling for answers. OpenAI confirmed both disruptions and rolled out fixes each time, but the rapid back-to-back incidents renewed questions about the platform’s stability.
Thousands Report Issues on February 3
The first outage struck during the afternoon of February 3, 2026, when reports of ChatGPT not working began flooding Down Detector, the outage-tracking website. Reports climbed past 12,000 in the United States before starting to fall once OpenAI began applying fixes.
OpenAI confirmed two distinct problems during the disruption: elevated error rates for fine-tuning jobs, and elevated errors affecting both ChatGPT and its broader developer platform. For everyday users, this meant the chatbot either fell silent or responded far more slowly than normal.
The impact was felt most sharply in the US and UK. Users in the UK found that connecting through a US-based VPN allowed them to access ChatGPT while the service remained unstable internationally. Testing also showed that VPN connections through Ireland, Canada, India, and Poland restored access, though India and Poland were somewhat slower to respond.
By 4:21 PM Eastern Time, OpenAI updated its official status page to confirm it had identified the problem and applied the necessary mitigations, adding that it was monitoring recovery. Just minutes later, Down Detector reports in the United States had dropped to roughly 333 — a clear sign that service had largely stabilized.
A Second Strike Within 24 Hours
Less than a day later, history repeated itself. Around 9:30 a.m. Pacific Time on February 4, 2026, ChatGPT began going down again. Reports on Down Detector shot past 10,000 within minutes, eventually peaking at just over 15,000.
The problems users described were wide-ranging. Projects wouldn’t load. Chat histories disappeared. Image uploads failed, with some users reporting a 403 error. And in many cases, the chatbot stopped responding to messages altogether. A user in Toronto shared screenshots showing the 403 error when attempting to load a profile and upload images, while similar complaints surfaced from Southern California and other locations.
One important detail: the February 4 outage was confined to ChatGPT itself. OpenAI’s other products, including its video-generation tool Sora and its developer APIs, remained fully operational throughout.
OpenAI’s Delayed Response, Fast Fix
When the February 4 outage began, OpenAI’s status page still showed everything in green — fully operational across the board. The company updated the page only after a wave of user reports made the disruption impossible to ignore, first acknowledging elevated errors, then confirming an active investigation was underway.
From there, things moved fairly quickly. OpenAI applied a mitigation and announced it was monitoring ChatGPT’s recovery. Reports on Down Detector fell below 500 within roughly half an hour of the peak, and the full disruption lasted approximately 35 to 45 minutes from its worst point to near-full resolution — a timeline that closely mirrored what unfolded the previous day.
A Pattern That’s Hard to Ignore
These two incidents didn’t come out of nowhere. In the weeks before February 3 and 4, ChatGPT had been experiencing a recurring series of short-lived outages, each following a familiar cycle: a sudden spike in user complaints, a brief acknowledgment from OpenAI, a fix applied, and then recovery in under an hour.
No major, multi-hour outages were reported during this same stretch, meaning ChatGPT avoided any truly catastrophic failures. But the frequent 35-to-45-minute disruptions tested the patience of regular users, especially those on paid plans like ChatGPT Plus or Pro who depend on the service for professional work and daily productivity.
OpenAI has not publicly explained what is driving these recurring interruptions. Whether the cause lies in infrastructure capacity, increasing user demand, or something else entirely remains unclear. What is clear is that as ChatGPT continues to attract more users and take on a larger role in how people work, the tolerance for repeated downtime — however brief — is shrinking fast.
