On June 10, 2025, a massive outage of ChatGPT, OpenAI’s flagship AI chatbot, left millions of users across the globe unable to access the service, affecting free accounts, premium subscriptions, and enterprise users alike. Starting around 1 a.m. EDT, the disruption triggered error messages like “something went wrong” and “too many concurrent requests,” alongside sluggish performance in models such as 4o-mini. Reports poured in from countries including the United States, Norway, New Zealand, and Indonesia, highlighting the scale of the issue. OpenAI acknowledged the problem, stating it is investigating, but has yet to provide a clear resolution. The outage also impacted Sora, OpenAI’s AI video generation tool, underscoring the world’s growing dependence on AI technologies. Users reported disruptions in professional tasks, academic work, and creative projects, with frustration mounting due to OpenAI’s limited communication.
The outage caught many off guard, given ChatGPT’s critical role in daily workflows. Social media platforms buzzed with complaints, and Downdetector logged over 1,200 reports by 7 a.m. EDT. The incident revealed how integral AI tools have become, from content creation to project management, while exposing vulnerabilities in their infrastructure.
- Affected models: 4o, 4.1, and core APIs, with frequent errors.
- Temporary fixes: Lower-tier models like 3o and 4.1-mini work, but slowly.
- Global reach: Complaints from Australia, Germany, Philippines, and beyond.
OpenAI’s vague updates fueled speculation, with no concrete timeline for a fix. The outage sparked discussions about the reliability of AI systems as their adoption continues to soar.
Scope of the disruption
The ChatGPT outage impacted users in at least 12 countries, with reports peaking in the early hours of June 10. Downdetector recorded over 1,200 complaints by 7 a.m. EDT, particularly from Europe and North America. Free account holders faced the most severe issues, encountering constant error messages, while ChatGPT Plus and Enterprise users reported significant slowdowns.
In Ohio, a user named Keith described struggles accessing the service on both the app and web, despite being a Plus subscriber. In Norway, Felix reported the “something went wrong” error appearing almost instantly after any interaction. In Indonesia, the 4.1 model failed entirely, though 4o-mini operated sluggishly.
Sora, OpenAI’s video generation tool, also suffered, with four components affected, per the company’s status page. The lack of clarity on the outage’s cause frustrated users, particularly those relying on these tools for time-sensitive tasks.
User experiences
The outage disrupted a wide range of users, revealing ChatGPT’s deep integration into daily life. Paul, a multimedia professional from New Zealand, said the failure halted his ability to calculate billable hours and manage workflows. He noted prior instability, with messages occasionally ignored by the system.
In Philadelphia, Gwen, a self-described power user, relied on ChatGPT for up to 16 hours daily and felt the outage keenly, citing latency and unresponsive queries. Megan, a student, said the failure disrupted exam preparation, as she and her peers leaned on the tool for study support.
- Disrupted tasks: Content creation, financial calculations, academic support.
- Usage frequency: From 2 to 16 hours daily, varying by user.
- Workarounds: Some switched to lower-tier models like 3o, with mixed results.
- Outage duration: Issues reported since early June 10.
These accounts highlight the widespread reliance on AI and the lack of immediate solutions from OpenAI, which has yet to pinpoint the technical cause.
Model performance
Not all ChatGPT models were equally affected. Users with access to 3o and 4.1-mini reported partial functionality, though responses were slower than usual. The popular 4o model, however, consistently failed, often displaying errors after one or two interactions.
Kevin from Ohio noted that 3o worked with tolerable lag, while 4.1-mini responded normally in some instances. The 4o-mini model, though operational, required patience due to delayed responses. Enterprise account holders, like a TechRadar senior editor, accessed the service but faced subpar performance.
The varying performance suggests the outage stems from server overload or issues with critical components of advanced models. OpenAI confirmed 21 ChatGPT components and 14 APIs were impacted but offered no further details.
Social media reactions
The outage sparked a flood of comments on platforms like Reddit and X, where users shared frustrations and sought solutions. A Reddit user described ChatGPT as “painfully slow” before it displayed red error messages, while another said the service “freaked out” during new conversations. Some speculated that OpenAI was rolling out a major update, though no evidence supports this claim.
Posts on X criticized OpenAI’s lack of transparency, with users demanding clearer updates. One post noted that even ongoing projects were affected, with generic prompts failing repeatedly. The absence of detailed official statements fueled discontent, especially among paying subscribers.
Temporary workarounds
Some users found partial solutions, though with caveats. Mitchell, a Plus subscriber, recommended switching to lower-tier models like 3o, which showed fewer errors. He also observed odd behavior, such as responses formatted in Canvas with code-like markings for simple queries.
Michael, another user, reported that existing project chats continued to function, albeit slowly, while new, generic prompts failed. This led some to focus on pre-existing chats to maintain limited access.
- Model switching: 3o and 4.1-mini as viable alternatives.
- Project focus: Existing chats more likely to respond.
- Limitations: Slow responses persist in workarounds.
Despite these efforts, most users remained without full access, awaiting a comprehensive fix.
History of instability
The June 10 outage wasn’t an isolated incident. Users reported intermittent issues in prior weeks, including unprocessed messages and slow responses. Paul from New Zealand mentioned occasional failures requiring multiple attempts to get a reply.
OpenAI has not confirmed whether these incidents are related, but their frequency raises questions about the scalability of its infrastructure. The demands of processing billions of daily interactions may strain servers, particularly as tools like ChatGPT and Sora grow in popularity.
Reliance on AI
The outage underscored ChatGPT’s centrality to modern workflows. From students like Megan, who used it for exam prep, to professionals like Daryl, who relied on it for YouTube thumbnail creation, the disruption halted diverse tasks. The tool’s role has expanded beyond casual chats to critical functions like multimedia archiving and financial calculations.
This widespread dependence highlights the risks of over-reliance on AI, especially when infrastructure falters. The outage disrupted entire workflows, emphasizing the need for robust systems as AI adoption accelerates.
Sora’s status
Sora, OpenAI’s video generation tool, also faced issues, with four components impacted, according to OpenAI. Users like Labi from Norway reported that Sora failed to load, compounding frustrations. The simultaneous failures of ChatGPT and Sora suggest a broader issue, possibly tied to shared servers or core systems.
The lack of specific updates on Sora deepened user dissatisfaction, particularly for those relying on it for visual content creation. OpenAI’s silence on the matter left creators in limbo.
Lack of official clarity
OpenAI’s updates remained minimal, stating only that “some users are experiencing elevated error rates and latency” and that investigations are ongoing. The latest update at 7:09 a.m. EDT offered no resolution timeline, drawing criticism from users like Dustin in Munich and Jonny, a free user, who faced generic error messages.
The company’s reticence clashed with the outage’s scale, which affected casual users and businesses using OpenAI’s APIs. TechRadar reached out to OpenAI for comment but received no further details.