ChatGPT experienced a service disruption on Tuesday evening, July 14, 2026, leaving thousands of users unable to log in, open conversations or receive responses. OpenAI confirmed that it was investigating login problems and intermittent errors affecting the AI chatbot.

User reports began increasing shortly after 7:30 p.m. ET. Some people were met with blank pages or failed login attempts, while others could access ChatGPT but could not submit prompts or load previous conversations.

The disruption was not affecting every account in the same way. Some users reported that the mobile app remained available while the website failed, while others experienced problems across both web and mobile platforms.

ChatGPT outage reports rose rapidly

Downdetector initially recorded nearly 2,000 reports connected to OpenAI shortly after the disruption began. The number then increased above 4,000, passed 5,000 and reached more than 7,000.

By approximately 5:18 p.m. PT, more than 10,000 users had submitted reports about problems with the platform. More than 80% of the reported OpenAI issues were linked to ChatGPT, according to figures shared during the incident.

Downdetector figures are based on reports submitted by users and do not represent the total number of people affected. However, a sudden increase in reports can help confirm that a problem is broader than one account, device or internet connection.

When did the ChatGPT outage begin?

• Shortly after 7:30 p.m. ET: Users begin reporting that ChatGPT is failing to load.
• Around 7:53 p.m. ET: Nearly 2,000 outage reports are recorded.
• 7:57 p.m. ET: OpenAI confirms it is investigating login issues and intermittent errors.
• 4:57 p.m. PT: Reports climb above 4,000.
• 5:07 p.m. PT: More than 7,000 reports have been submitted.
• 5:18 p.m. PT: The total rises above 10,000 reports.

The Eastern and Pacific times refer to the same developing incident. Different publishers reported the timeline using their local time zones.

What error message were users seeing?

Some affected users reported seeing the following message when attempting to access ChatGPT:

“upstream connect error or disconnect/reset before headers. retried and the latest reset connection timeout.”

This type of error generally means that one server or part of the service could not connect successfully with another system needed to process the request.

It does not necessarily mean that a user’s account has been suspended, hacked or deleted. Other reported problems included blank screens, login failures, conversations that would not load and prompts that remained unanswered.

What did OpenAI say?

OpenAI acknowledged the disruption in a 7:57 p.m. ET update, stating that it was “investigating login issues and intermittent errors affecting ChatGPT.”

The notice confirmed that at least part of the problem was occurring within OpenAI’s service rather than being caused by an individual user’s browser or Wi-Fi connection.

OpenAI had not publicly identified the technical cause or provided an estimated restoration time during the initial investigation. Users can monitor the official OpenAI status page for updates on whether the company is investigating, applying a fix, monitoring recovery or has resolved the incident.

Was ChatGPT completely down?

No. OpenAI described the problem as involving intermittent errors, which means service availability could vary between users, regions and platforms.

Some people could open the website but could not send messages. Others could not log in at all. A smaller number reported that the ChatGPT app continued working even though the web version was unavailable.

The reports did not establish that the incident was limited to free accounts. Login and conversation-loading problems may also affect Plus, business and workplace users when the disruption involves systems shared across subscription plans.

Is ChatGPT down or is it a device problem?

When thousands of users report similar errors within minutes and OpenAI confirms an active incident, the problem is unlikely to be caused by one device or home internet connection.

Users can try opening another website, switching between the app and web version or using a private browser window. However, clearing cookies, reinstalling the app or resetting a password will not fix a platform-wide outage.

This is not the first time users have faced access problems. A previous ChatGPT loading outage also prevented users from accessing the service, while another incident saw outage reports surge among thousands of ChatGPT users.

Will users lose their ChatGPT conversations?

A temporary failure to display conversation history does not automatically mean that saved chats have been permanently deleted. Conversations may disappear temporarily or fail to load while ChatGPT systems are experiencing errors.

Users should wait until OpenAI confirms that service has recovered before assuming that content has been lost. Important responses, drafts and research should still be copied into a separate document whenever possible.

What should users do while ChatGPT is unavailable?

Users should avoid repeatedly refreshing the page or submitting the same prompt several times. Duplicate requests may process after the service begins recovering.

Trying the mobile app or returning later may help, but there is no guaranteed user-side fix during a confirmed OpenAI incident. People working against deadlines should temporarily use another writing, research or coding workflow.

Users should also be cautious of unofficial links claiming to restore ChatGPT access. Outages can be used to promote fake login pages or phishing tools designed to collect account information.

Read More:

• ETIAS delayed to 2026: What the €20 EU travel authorisation means for travellers
• Qantas $105 million class action settlement: Who can claim and what it means
• Visit the Swikblog homepage for more latest news

Why the outage matters

ChatGPT is widely used by students, teachers, developers, researchers, journalists, marketers and customer-support teams. Even a relatively short interruption can delay assignments, coding work, document preparation and automated business tasks.

The July 14 disruption highlights the risks of relying on one cloud-based AI service for time-sensitive work. OpenAI continued investigating the incident without initially providing a confirmed cause or timetable for full recovery.