OpenAI Outage Watch: Tracking Every Glitch Since GPT-5
Since GPT-5’s debut, OpenAI’s services have faced recurring disruptions from elevated error rates in flagship models to outages in search, file handling, and payment systems.
What began as an unstable launch week has evolved into an ongoing operational challenge. This page now functions as a live OpenAI incident tracker updated daily with verified reports from OpenAI.
Live OpenAI Incident Log
• Codex, Connectors, Search, Agent, Deep Research, Custom GPTs, Project Sharing, ChatGPT Record, Memory and other features related to RBAC are missing, unavailable, or unexpectedly enabled even when toggled off for Enterprise and Edu workspaces (Tag: Minor Outage)
From 20:37–17:28 UTC, Codex, Connectors, Search, Agent, Deep Research, Custom GPTs, Project Sharing, ChatGPT Record, Memory and other features related to RBAC are missing, unavailable, or unexpectedly enabled even when toggled off for Enterprise and Edu workspaces
• Increased errors on Google signup for Android users (Tag: Resolved – Degraded Performance)
Android users experienced elevated errors when using Google Sign-In for ChatGPT. OpenAI investigated, applied mitigation, then monitored recovery and declared full service restoration once success rates returned to baseline.
• Compliance API ingestion delays (Tag: Resolved – Degraded Performance)
Compliance API ingestion suffered data delays from Sep 29. OpenAI investigated, applied mitigation, monitored for a day, and confirmed full recovery, with the Compliance API back to normal functioning.
• Increased error rate for ChatGPT (Tag: Resolved – Degraded Performance)
ChatGPT showed elevated error rates during the afternoon. OpenAI investigated, mitigated, then monitored improvement and marked the incident resolved, confirming all impacted services fully recovered.
• Increased latency on fine-tuning jobs (Tag: Identified)
Fine-tuning jobs took longer than expected due to high load beginning 16:55 UTC. OpenAI identified the cause and started mitigation, continuing to monitor job completion times while capacity normalized.
• Elevated errors for Code Interpreter in the API (Tag: Identified)
The Code Interpreter endpoint in the API showed elevated errors from 10:16 UTC. Engineers identified the issue and began mitigation, with recovery updates to follow as stability returned.
• Elevated errors in Conversations with ChatGPT (Tag: Resolved)
From 11:59–13:54 UTC, some ChatGPT conversations failed, especially for logged-out users. OpenAI implemented a mitigation, monitored improvement, and confirmed full recovery with success rates back to baseline.
• ChatGPT on WhatsApp outage (Tag: Identified)
Starting 03:05 UTC, ChatGPT on WhatsApp returned elevated errors. OpenAI identified the issue by 04:25 UTC and worked on mitigation as recovery efforts continued.
• ChatGPT login degradation (Tag: Resolved)
Between 21:23–21:47 UTC, users experienced login issues in ChatGPT. The team identified the fault, applied mitigation, monitored recovery, and marked the incident resolved the same hour.
• Increased errors while using Connectors in ChatGPT (Tag: Resolved)
Connectors in ChatGPT intermittently failed from 19:16–20:25 UTC. OpenAI mitigated the issue and confirmed recovery after success rates stabilized.
• Increased error rate for ChatGPT (Tag: Investigating)
At 14:58 UTC, OpenAI reported a partial outage with increased ChatGPT errors across components. The team investigated and issued updates as mitigation work progressed.
• ChatGPT: elevated errors on file uploads (Tag: Identified)
From 04:18 UTC, some users saw file-upload errors in ChatGPT. OpenAI identified the problem by 08:32 UTC and began mitigation, continuing to monitor recovery.
• ChatGPT safe tables delay 7days 09/22 (Tag: Monitoring)
OpenAI noted a delay affecting ChatGPT safe tables on 09/22. Mitigation was applied and monitoring continued while data pipelines caught up and validations passed.
• Users unable to create, edit, or upload files to projects — resolution (Tag: Resolved)
An incident spanning Sep 18–20 UTC impacted project file create/edit/upload operations. Mitigation concluded and full recovery was confirmed at 00:12 UTC on Sep 21, restoring project workflows.
• Users unable to create, edit, or upload files to projects (Tag: Identified)
Beginning 19:53 UTC, users couldn’t create, edit, or upload files to projects. OpenAI identified the issue by 21:13 UTC, applied mitigation at 23:10 UTC, and continued monitoring until recovery.
• Elevated error rates on ChatGPT Agent & GPT-5 Pro (Tag: Investigating)
At 09:49 UTC, OpenAI began investigating elevated error rates affecting ChatGPT Agent and GPT-5 Pro. Mitigation steps followed, with updates posted as recovery progressed.
• Errors for Conversations with Images for Free Users (Tag: Identified)
From 15:03 UTC, free-tier users hit elevated errors when sending images in ChatGPT conversations. OpenAI identified the issue at 15:05 UTC and began mitigation.
• Elevated errors with SSO login (Tag: Monitoring)
Starting 16:47 UTC, SSO logins produced elevated errors. OpenAI applied mitigation by 17:06 UTC and monitored recovery until stability was confirmed.
• Elevated errors in GPT Store (Tag: Identified)
At 15:43 UTC, the GPT Store experienced elevated errors. OpenAI identified the fault by 16:11 UTC and worked on mitigation to restore storefront reliability.
• Users may see both a GPT-5 response and an agent response when running agent tasks (Tag: Resolved)
Some agent runs produced two responses—one from GPT-5 and one from the agent—causing confusion in results. Mitigation was applied and recovery confirmed, with behavior returning to normal.
• Impact in retrieval quality on file_search (Tag: Monitoring)
Retrieval quality is degraded for files added Sep 10 (PT). Workaround: remove and re-add those files to vector stores. OpenAI is monitoring recovery and continuing to fix impacted files.
• Increased number of users reporting rate limit errors with GPT-4o model (Tag: Resolved)
Some ChatGPT users hit rate limits earlier than expected on GPT-4o. A mitigation was deployed, recovery monitored, and service returned to expected rate-limit behavior shortly after.
• Deep Research elevated errors (Tag: Resolved)
Deep Research showed elevated errors for a subset of users. Engineers implemented a mitigation, monitored improvement, and marked the incident resolved once success rates stabilized.
• Fine-tuning jobs experiencing delays (Tag: Resolved)
New fine-tuning jobs stalled or took longer to finalize. Inference on existing fine-tuned models was unaffected. OpenAI implemented a fix, processed delayed jobs, and confirmed resolution.
• Errors with Batch API (Tag: Resolved)
The Batch API returned elevated errors for requests. OpenAI identified the issue, implemented mitigation, monitored recovery, and confirmed full restoration of normal batch processing behavior.
• File Search not working across Responses and Assistance API (Tag: Resolved)
file_search failed across Responses and Assistants APIs for some users. Engineers investigated, deployed a mitigation, and confirmed services were fully recovered with normal retrieval behavior restored.
• ChatGPT Agent elevated error rate (Tag: Resolved)
The ChatGPT Agent experienced elevated error rates, affecting execution reliability for agent tasks. OpenAI implemented mitigation, monitored stability, and confirmed normal success rates after resolution.
• Increased error rates in Web Search tools in Responses API (Tag: Resolved)
Web Search tools within the Responses API produced elevated errors, impacting result retrieval. OpenAI identified the cause, deployed mitigation, and restored normal behavior, closing the incident as resolved.
• Increased error rates in Web Search tools in Responses API (Tag: Minor Outage)
From 09:32–09:47 UTC, Increased error rates in Web Search tools in Responses API
• ChatGPT Not Displaying Responses (Tag: Partial Outage)
From 03:10–04:13 UTC, ChatGPT responses failed to render for some users. OpenAI identified the root cause and is working to resolve the partial outage, with recovery updates pending as monitoring continues.
• Elevated errors on ChatGPT (Tag: Resolved)
Between 10:23–10:41 UTC, ChatGPT experienced elevated error rates affecting conversation reliability. OpenAI applied mitigation, monitored stability, and confirmed full recovery, with success rates returning to baseline.
• Elevated errors on ChatGPT (Tag: Resolved)
Users experienced elevated errors in ChatGPT starting at 12:26 UTC. OpenAI identified the issue, applied a mitigation by 01:00, and confirmed full recovery at 02:03 UTC, with services returning to normal performance.
• Increased login and signup errors (Tag: Resolved)
Logins and signups for ChatGPT and Platform failed intermittently from 14:14–15:25 UTC. OpenAI identified the issue, deployed mitigation, monitored recovery, and reported full restoration, with users no longer experiencing authentication errors.
• Search in ChatGPT failing for some free users (Tag: Resolved)
For free-tier users, Search in ChatGPT intermittently failed between 19:58 and 00:51 UTC (Aug 28). OpenAI identified the fault, applied mitigation, monitored recovery, and declared full recovery once success rates normalized.
• Elevated error and latency for API (Tag: Resolved)
APIs experienced elevated errors and latency beginning 21:01 UTC, with mitigation and monitoring overnight. OpenAI confirmed full recovery at 03:47 UTC on Aug 27, with request success rates and response times back to baseline.
• Elevated errors in ChatGPT Voice (Tag: Resolved)
Voice mode conversations in ChatGPT could not start or failed to connect for a subset of users from 10:13–11:00 PT. OpenAI resolved the incident and restored normal connection reliability shortly afterward.
• Voice conversations dropping previous messages (Tag: Resolved)
Voice conversations in ChatGPT dropped previous messages mid-call, breaking context for some users. Engineers identified the issue, applied mitigation, and confirmed full recovery by 02:15, restoring conversation history stability.
• Elevated API errors and latencies (Tag: Resolved)
OpenAI’s APIs saw elevated errors and latency across multiple components. Teams identified the cause, applied mitigation, and confirmed full recovery by 23:12, with request success rates and response times returning to baseline.
• Increased latency on realtime, chat completions and text-embedding APIs (Tag: Resolved)
Increased latency hit realtime, chat completions, and text-embedding APIs. Partial recovery rolled out during investigation; mitigation completed and full recovery was confirmed by 23:06 after monitoring restored expected performance across affected endpoints.
• Sora experiencing elevated error rates and increased latency (Tag: Resolved)
Sora experienced elevated errors and latency, temporarily impacting generation workflows. Mitigation was deployed and full recovery confirmed by 19:00, with continued monitoring to ensure stability across affected components.
• ChatGPT Conversation Errors (Tag: Resolved)
Some ChatGPT Conversations degraded with elevated errors. OpenAI mitigated the issue, monitored recovery, and marked the incident resolved at 18:30, restoring normal conversation reliability for impacted users.
• Help Center login is down (Tag: Resolved)
Help Center logins failed for some users. A mitigation was implemented, recovery was monitored, and full restoration confirmed at 17:21, returning access to support resources and account management.
• “Failed To Load Subscription” issue (Tag: Resolved)
Users saw “Failed to load subscription” errors in ChatGPT. OpenAI stabilized entitlements with mitigation, monitored recovery, and declared the issue fully resolved at 15:09, restoring subscription visibility and billing access.
• ChatGPT elevated error rate (Tag: Resolved)
ChatGPT returned elevated error rates across many components early morning. Sequential mitigations were applied and full recovery confirmed by 07:58, with normal success rates observed after monitoring.
• Elevated Errors for File Uploads to Vector Stores (Tag: Resolved)
File uploads to vector stores reported elevated errors. Investigation began, mitigation was applied, and full recovery confirmed at 21:42, restoring ingestion pipeline performance for affected customers.
• GPTs with custom actions stuck processing (Tag: Resolved)
Some GPTs with custom actions stalled in a processing state. A temporary Privacy settings workaround helped some users while mitigation was implemented; recovery was monitored and resolution confirmed at 21:14.
• Errors logging into ChatGPT with Apple (Tag: Resolved)
Users encountered elevated errors signing into ChatGPT with Apple from 10:50–11:36 PT. The issue was quickly investigated and resolved shortly afterward, with normal sign-in flows and account access restored.
• Images not loading on the iOS, Android and macOS apps (Tag: Resolved)
On iOS, Android, and macOS apps, images failed to load for hours. OpenAI investigated, applied mitigation at 18:01, and confirmed full recovery at 23:41, restoring media rendering across platforms.
• Elevated Error Rates on ChatGPT Agent (Tag: Resolved)
ChatGPT Agent experienced elevated error rates into early morning. Mitigation restored stability; OpenAI monitored and confirmed full recovery at 02:57, after errors began late on Aug 15.
• Increased errors in Chat Conversation (Tag: Resolved)
ChatGPT conversations produced increased errors between 01:20–02:19. A quick mitigation restored service, with recovery monitored and the incident closed once success rates returned to expected levels.
• Codex Errors (Tag: Partial Outage)
Codex-related operations in ChatGPT saw elevated errors between 00:14–01:06 UTC. Mitigation was applied, and OpenAI confirmed full recovery shortly after.
• ChatGPT Connector Errors (Tag: Degraded Performance)
Plus and Pro users encountered elevated errors with the “Use connectors” tool, which was inaccessible for some. Issue identified and mitigation in progress at the time of the update.
• Increased Errors on Compliance API (Tag: Partial Outage)
The Compliance API experienced elevated errors in the evening. A mitigation was applied, recovery was monitored, and services were marked fully recovered by 19:27 UTC.
• Increased Errors in Image Generation (Tag: Degraded Performance)
ChatGPT image generation returned elevated errors from roughly 06:40–07:22 UTC. Recovery was monitored, and the incident was marked as resolved soon after.
• Image Generation Errors for Free ChatGPT Users (Tag: Degraded Performance)
Only free-tier ChatGPT users were affected by elevated image-generation errors around 12:06 UTC. A mitigation was implemented and the incident was resolved by 12:20 UTC.
• Partial Search Outage for ChatGPT Users (Tag: Partial Outage)
Web search via API and ChatGPT prompts triggering searches suffered increased latency. A fix was deployed and normal functionality restored by 22:02 UTC.
• Increased Failure Rate for Codex Tasks (Tag: Partial Outage)
Codex task execution saw higher failure rates between 19:56–22:02 UTC. A fix was implemented, monitored, and normal functionality restored.
• Rate Limit & “Model Not Found” Errors (GPT-5) (Tag: Degraded Performance)
Users faced rate limit and “model not found” errors in GPT-5 sessions for over an hour. Mitigation restored normal service by evening.
• Elevated Error Rates in GPT-5 Conversations (Tag: Partial Outage)
Late-night incident caused elevated error rates in GPT-5 chats, affecting free and paid users. Services were restored within hours; RCA promised.
• Deep Research (Free Users) Elevated Errors (Tag: Degraded Performance)
Free-tier users encountered elevated errors in Deep Research tools. Engineers investigated promptly, with fixes rolled out the same day.
• 502 Gateway Errors (ChatGPT) (Tag: Partial Outage)
Widespread 502 Gateway errors affected ChatGPT throughout the day. Mitigation restored full service by evening.
• Elevated Errors in ChatGPT Conversations (Tag: Degraded Performance)
Conversation sessions in ChatGPT saw error spikes on the evening of August 6. The incident was resolved in hours; RCA pending.
Launch Week Summary & GPT-5 Rollout Analysis
The rollout of GPT-5 brought notable instability across OpenAI’s ecosystem, hitting models, search, file handling, and payments during its first days.
Elevated Errors in GPT-5 and GPT-4.5
On August 8, GPT-5 conversations faced elevated error rates, disrupting chats and causing incomplete responses. Earlier that day, GPT-4.5 experienced similar issues, suggesting shared backend strain.
Search Downtime and File Handling Issues
Search functionality was partially unavailable, limiting external and historical data retrieval. File uploads to vector stores and Google Drive imports suffered latency, delaying AI-powered document processing.
Payment and Agent Service Disruptions
Credit card transactions for new API sign-ups failed briefly. Plus and Pro subscribers had ChatGPT Agent feature issues during the launch window.
Continuing GPT-5 Issues
On August 9, rate limit and “model not found” errors returned, marking the second day of GPT-5-specific disruptions. Though resolved quickly, it underscored the strain of rapid adoption.
Recovery and Next Steps
Most incidents were resolved within hours, but the breadth of impact highlighted scaling readiness concerns. The Root Cause Analysis will determine if these were launch-phase turbulence or signs of deeper infrastructure bottlenecks.
Conclusion: Why This Tracker Matters
The launch of GPT-5 has been both a milestone in AI capability and a stress test for OpenAI’s infrastructure.
While recovery times are improving, service stability and load management will be as critical as the technology itself.
By maintaining this live tracker, we aim to give users a clear, timely, and structured view of issues as they unfold, ensuring transparency beyond the initial hype cycle.
For more AI stories, visit AI News on our site.
