ChatGPT Down for Over 12 Hours: What Happened on June 10, 2025?

 You’re likely not going to see AI assistants take a day off and relax, but on June 10, 2025, that’s exactly what happened. OpenAI’s popular chatbot, ChatGPT, suffered a global outage for 12 hours straight, creating a frenzy and panic worldwide, all while memes were being created. In this report I will go over the timeline, impact, causes, and the important lessons that can be taken away from this domino effect event.

Timeline: When did the problems begin?

• 2:45 AM ET (12:15 AM PT): Users from North America, Europe, and Australia are unable to connect with ChatGPT and Sora. There is a spike in outages reported by Downdetector.  
• 10 AM ET: OpenAI confirms the “elevated error rates and latency” they were facing and has identified a solution that is in progress.  
• 6:32 PM ET: ChatGPT displays API access and improvements, which are sustained although voice interactions and features still remain unsteady.  
• 11:55 PM IST: The outage persists in India, thereby calculating over 12 hours of disruption worldwide, which is likely the longest downtime in the history of ChatGPT.

Who faced the consequences?  

This was not a small inconvenience; it was a catastrophic breakdown.

• Affected by region: Participants from India, US, UK, Europe, Australia and other regions suffered outage related issues that included ChatGPT, Sora and APs.
• Service disruption: The outage affected all users: both free and paid subscribers, API use, GPT-4o feature, voice chat, as well as third party Perplexity services.  
• User frustration: Users faced prompts with errors such as “Hmm… something seems to have gone wrong” and “A network error occurred.” Nearly all features apart from the separate playground were impacted.  

The Internet Reacts  

With AI tools disabled, the internet went full meme:  

• “Am I meant to use my brain at work now?”  
• “ChatGPT is down and I’m mid-flight piloting a plane—send help!”  

This incident also served to emphasize the extent to which people depend on AI for routine activities.  

Why It Matters  

OpenAI admitted that a sudden surge in use had strained its infrastructure, with some components reporting capacity limits. Engineers reported that “GPUs are melting” due to demand.  

• Global ripple impacts: Students, professionals, and developers around the world suddenly lost access to critical resources such as writing, coding, and research tools.  
• Trust & reliability: Organizations that depend on 99.9% uptime from AI services for business operations took this outage as an eye opener regarding resiliency and over-dependency.

How OpenAI Responded

• Communications: As an example, the company resolved updates as they worked through the issue, posting “We have identified the root cause…mitigation in progress.” 
• Recovery in phases: API Services recovering first at around 6:30 PM ET, followed by the features of ChatGPT and voice services recovering subsequently.
• Investigative follow up: OpenAI confirmed the cause was resolved and all services functioned fully by the late evening.

What You Can Learn from This

• Learn to Explore: Always enable backup services like Grok, Gemini, Claude, Bing AI, or Perplexity. 
• Users should be kept informed: If your services and applications are powered by ChatGPT, Comprehensive communication is mandatory. 
• Relying on AI requires building systems with resilient architecture: Design the workflows that are critical and reliant on AI to ensure they are able to withstand or degrade steadily during outages.
• Learn from every incident: Investigate capacity limits, sudden demands or bugs; find ways to reinforce the infrastructure.

In Summary

The June 10, 2025 outage is marked with the most significant ChatGPT interruption, 12+ hours of widespread global downtime affecting millions. It highlighted the importance of AI’s role in our daily lives, as well as revealing crucial weaknesses in the systems and dependencies, infrastructure. Striving to improve systems, Open AI is reminded that in a framework dependent on AI, the systems must be designed to fail regardless of the innovations made.