AI Goes Dark: ChatGPT Meltdown Sends Millions Scrambling!
Glasgow, Scotland – June 10, 2025 – OpenAI, the leading force behind popular AI tools like ChatGPT and Sora, is currently navigating a significant global outage that has left countless users frustrated and workflows disrupted. This latest incident, which began in the early hours of June 10th, marks another chapter in a turbulent four-month period characterized by frequent and widespread service interruptions across the company's expanding suite of artificial intelligence products.
Today's outage, initially reported on outage tracking sites like Down Detector around 7:45 AM BST, quickly escalated, affecting users across the UK, Europe, North America, India, and Australia. By mid-afternoon, OpenAI's status page confirmed "elevated error rates and latency" across its core services, including ChatGPT, the API, and the text-to-video generator Sora. While an update around 8:06 AM BST hinted at recovery for API calls, full restoration for all affected services, particularly ChatGPT, is still expected to take "another few hours."
Users attempting to access ChatGPT have been met with various frustrating symptoms: the chatbot failing to load, displaying "Hmm…something seems to have gone wrong" or "A network error occurred" messages, or even exhibiting "flat, no personality, even fabricated memories of conversations and total chaos in replies," as one user described. Reports indicate that models like ChatGPT o3 were particularly affected, though even some 4o-mini users reported issues. The disruption has impacted both free and paid-tier users, with a staggering 93% of complaints on Down Detector related to ChatGPT simply not responding.
A Pattern of Instability: March to June 2025
This isn't an isolated event. The past four months have seen a remarkable frequency of disruptions, raising concerns about the reliability of an infrastructure that is becoming increasingly vital for daily tasks and business operations.
June 2025's Rocky Start: Beyond today's major outage, June has already seen:
Elevated error rates for omni-moderation-latest (June 6th).
Elevated error rates for file uploads in ChatGPT and API (June 5th).
The ChatGPT Windows client not responding and Deep Research unable to access content from uploaded files (June 4th).
Compliance API Data Delays (June 3rd) and issues with exporting user analytics data and increased errors rates on gpt-4.1-nano (June 2nd).
May 2025: A Month of Minor Meltdowns: May proved to be a particularly challenging month, with a broad range of OpenAI's offerings facing issues:
Usage cap errors in ChatGPT (May 31st).
Increased errors for ChatGPT conversations and connectors, including an "Error fetching connectors connections" message (May 29th).
Rising error rates in Codex, Deep Research, and ChatGPT Login (May 28th).
Codex unavailability (May 27th) and subsequent task and PR creation errors (May 26th).
A significant incident where some ChatGPT Pro and Team customers lost subscription access between May 22nd and May 26th due to billing renewal issues.
Multiple occurrences of elevated error rates in ChatGPT (May 22nd, May 13th) and Codex (May 23rd, May 22nd), with one Codex issue attributed to GitHub API downtime.
ChatGPT Login issues (May 21st) and increased errors for voice transcription (May 20th).
Problems with buying API credits, updating limits, and adding or updating payment methods for API customers (May 18th, May 17th).
ChatGPT search experiencing elevated failures and high latency (May 17th).
Image Generation API Errors with the
gpt-image-1model (May 15th).
April 2025: A Consistent Stream of Snags: April was far from smooth sailing:
Degraded Performance: Increased timeouts on gpt-4.1-nano and Sora experiencing elevated error rates (April 30th).
Login Issues Impacting Some Users and Degradation on Response API (April 30th).
Temporary delays in video generation capabilities for new users (April 29th).
Recurring increased error rates for Deep Research in ChatGPT (April 29th) and general increased error rates in ChatGPT (April 29th, April 28th).
A notable outage where platform.openai.com and ChatGPT were entirely unavailable (April 24th).
ChatGPT Login/Signup failure and Elevated Conversation Error Rate on Operator (April 23rd), prompting commitments for Root Cause Analyses (RCAs).
Elevated errors with Custom GPTs, Projects, and the Realtime API (April 22nd).
Elevated error rates for Text to speech (tts-1) (April 19th) and ResponsesAPI Image Input Errors (April 17th).
Increased error rates in Moderation API and DALL·E API (April 9th).
March 2025: The Genesis of Instability: The pattern of frequent disruptions arguably began in March:
Increased Error Rates in ChatGPT (March 31st) and File Uploads Errors (March 29th).
A significant external factor: a Cloudflare outage on March 27th, which, while not an OpenAI internal issue, caused high latency and timeouts for some OpenAI API customers due to Cloudflare's role in routing requests.
Degraded o3-mini function calling and ChatGPT Search not working with o3-mini (March 27th).
High error rate in Chat Completions API (March 26th) and Compliance API error rates (March 26th).
Image generation in Sora experiencing issues (March 26th).
Elevated 429 error rates (March 25th) and increased error rates in ChatGPT models 4.5 and 4o mini (March 25th).
A specific chat completions endpoint error on March 20th, caused by a faulty code deployment in a single cluster, led to a staggering 70% error rate for affected users.
While OpenAI's status page typically confirms that issues are being "investigated" and "mitigation" is being implemented, the lack of detailed root causes for many incidents leaves users in the dark. Some users, particularly paying subscribers, are questioning the value of their subscriptions when faced with such inconsistency. The recurrence of these outages also sparks broader discussions about the stability of cutting-edge AI infrastructure and the potential vulnerabilities that come with an increasing reliance on centralized AI services.
