ChatGPT / OpenAIIssue affecting voice mode in ChatGPT
All impacted services have now fully recovered.
Resolved
Outages and recoveries from the services you depend on.
40 incidents
All impacted services have now fully recovered.
Resolved
On July 13 between 15:30 UTC and 22:40 UTC, we experienced delays starting jobs on Docker (Gen 2) resource classes due to capacity constraints from our cloud provider. The same issue recurred on July 14 between 15:45 UTC and 22:50 UTC. Delays reached up to 25 minutes on July 13 and up to 24 minutes on July 14. The medium+ and 2 X-large+ resource classes saw the longest waits. Earlier updates in this incident reported figures based on average wait times, which didn't reflect the maximum impact some customers may have experienced. Wait times have now returned to normal. We are continuing to work with our cloud provider to add capacity ahead of the next peak period. Thank you for your patience.
Resolved
This incident has been resolved.
Resolved
## Summary From 01:33 UTC to 05:18 UTC on July 14, 2026, customers with a Bitbucket-linked identity were unable to log in to CircleCI, and some customers experienced workflows that failed to start or update, due to a change in how Bitbucket's OAuth service reports account permissions. At 05:07 UTC, we deployed a fix that corrected how our systems read the updated permissions field. Some customers needed to log out and re-authenticate their Bitbucket integration after we deployed the fix. Our systems continued processing the backlog of affected jobs until 08:39 UTC. We thank our customers for their patience as we resolved this incident. Please reach out to our support team with any questions or concerns. The status page for this incident can be found [here](https://status.circleci.com/incidents/gsyjwybg477g). ## Background CircleCI supports logging in with a GitHub account, a Bitbucket account, or an email and password. When a customer logs in with a Bitbucket-linked identity, or when CircleCI needs to refresh the customer's access on their behalf, we exchange an authorization token with Bitbucket's OAuth service. This exchange includes a list of the permissions, or "scopes," the customer has granted us, which Bitbucket and CircleCI use to confirm what CircleCI is authorized to do on the customer’s behalf. ## What Happened \(All times UTC\) On April 8, 2026, [Bitbucket announced a change to its OAuth service](https://developer.atlassian.com/cloud/bitbucket/changelog/#CHANGE-3139): it would be renaming the field used to report a customer's granted permissions. Bitbucket ran a transition period during which both the old and new field names were available, then fully removed the old field name on May 4, 2026. Our systems had not been updated to recognize the new field name, so once Bitbucket fully phased out the old one, requests that depended on it began to fail. At 01:33 on July 14, 2026, our systems began failing to process permissions information returned for Bitbucket-linked accounts, because our systems still expected the old permissions structure. This caused login attempts for all Bitbucket-linked accounts to fail, including for customers who log in with GitHub but have a Bitbucket identity linked to their account. At 01:59, automated monitoring alerted our engineering team to a spike in errors on the affected systems. The team began investigating immediately, confirmed customer impact at 02:38, and alerted customers via our status page at 02:51. By 03:00, the team had isolated the failures to the renamed Bitbucket field. Beginning at 03:17, workflow status updates for Bitbucket pipelines belonging to customers with an expired access token began to be dropped. The Bitbucket permissions check is used broadly across our platform, and these permissions failures also affected some of our internal job-processing systems. This caused a subset of workflows to become stuck without a final status, and caused some pull requests to show missing or stuck status checks. At 04:21, the team deployed an initial fix that resolved the underlying issue, restoring the login flow and returning our internal permissions system to normal operation. Some affected customers needed to log out and log back in to pick up the fix. By 05:07, the team deployed two additional fixes so our systems would begin accepting Bitbucket's new permissions field format going forward. We resolved the incident at 05:18. Some customers did need to log out and re-authenticate their Bitbucket integration before their account fully recovered. Our systems continued processing a backlog of affected jobs, returning to normal levels by approximately 08:39. ## Future Prevention and Process Improvement We are taking the following steps to prevent a recurrence and improve our response time: **We are hardening our authorization code against upstream API changes.** This incident happened because our system did not gracefully handle a renamed field in a response from a third-party identity provider. We will be immediately updating our authorization code so that unexpected or missing fields from GitHub, Bitbucket, and GitLab are handled safely. **We are improving how we track upstream provider changes.** We currently already monitor changelogs from our identity providers for exactly this kind of breaking change, but we didn't add this particular provider change to our monitoring in time to catch it before it shipped. We are auditing and expanding this monitoring so provider announcements reach our team before they affect customers. **We are improving how we classify and communicate incidents in their earliest minutes.** This incident's initial classification did not immediately reflect its customer-facing severity. We are refining our incident tooling and guidance to help engineers identify and communicate customer impact more quickly. **We are improving the resilience of our workflow-processing pipeline.** A downstream failure in this incident caused some workflow status updates to be dropped rather than retried or clearly surfaced. We are reviewing this system's retry and error-handling behavior so similar downstream failures are more visible and easier to recover from. Customer experience is our top priority, and we commit to continually improving the reliability of our systems to match the trust that our customers place in us. Please reach out to our support team with any questions or concerns.
Resolved
We've taken steps to mitigate the Speech Engine failures and have confirmed that performance has returned to baseline.
Resolved
Our Engineers have confirmed again that service has been successfully restored to all affected sites in the Amsterdam (NL) data center. This incident is now fully resolved.
Resolved
We resolved the performance bottleneck. All APIs are responding as expected as of now.
Resolved
A fix has been deployed, and all systems are performing normally. This incident has been marked as resolved.
Resolved
All services remain stable. The incident is now resolved.
Resolved
This incident has been resolved.
Resolved
Errors and latencies have recovered, incident has been resolved.
Resolved
Feedback ingestion is back to realtime and the backlog has been processed.
Resolved
The Help Center issues have been resolved by the third-party app and is fully operational again. Thank you for your patience!
Resolved
Help Scout Docs is fully operational - we’re closing this out. Thank you for your patience while we worked on this one!
Resolved
An effective fix for this issue has been applied and this incident is now resolved.
Resolved
This incident has been resolved. If you are still experiencing an issue, please let us know in our Etsy Forums (https://community.etsy.com/t5/Technical-Issues/bd-p/bugs).
Resolved
This incident has been resolved.
Resolved
This incident has been resolved.
Resolved
This incident has been resolved.
Resolved
We have resolved a performance issue that was causing elevated latency for some users in Latin American regions between July 3rd and July 14th, 2026. Our infrastructure team identified the root cause and the network routing configuration has been corrected. Traffic has been restored to the optimal path and performance has returned to expected levels. We sincerely apologize for the inconvenience and thank you for your patience.
Resolved
Services have become fully operational and the backfill for the impacted SBOMs have completed. We are continuing to monitor the platform to ensure stability.
Resolved
All impacted services have now fully recovered.
Resolved
This incident has been resolved.
Resolved
Error rates for room service api requests in our London region have remained normal since 09:06 am UTC and we are no longer seeing any issues.
Resolved
This incident has been resolved.
Resolved
The issue has now been resolved, and User Reports and Team Reports are loading as expected.
Resolved
After monitoring the situation, our Engineers have confirmed that full service has remained to our control panel, MyKinsta. If you are still unable to log into MyKinsta, please contact our support team by sending an email to support@kinsta.com. We do apologize for any disruptions caused.
Resolved
Issue with LinkedIn connections was fully restored.
Resolved
This issue is now resolved.
Resolved
This incident has been resolved.
Resolved
We experienced an elevated level of API errors and latencies for 3 minutes from 09:55 to 09:57 UTC. This incident has been resolved.
Resolved
Services have recovered and we are continuing to monitor the platform to ensure stability.
Resolved
Due to a change in security parameters, some customers are experiencing failures when attempting to sign in to Cloudinary via the Cloudinary Figma plugin (https://www.figma.com/community/plugin/1451492041396730459/cloudinary)
Resolved
This incident has been resolved.
Resolved
This incident has been resolved.
Resolved
This incident has been resolved.
Resolved
Our team has resolved the issue impacting some customers experiencing slowness or errors when accessing invoicing in Xero. We apologise for any inconvenience this has caused. Our technical team identified this was caused a momentary period of issues with network connectivity, and our team has restored service.
Resolved
The latencies were caused by issues with a downstream service which resolved after we scaled up. We've been monitoring for errors and latencies and have not seen any further issues. This issue is now considered resolved.
Resolved
A Java SDK release was incorrectly tagged, preventing the time-skipping test server from being published. After correcting the tag, rerunning the release workflow, and clearing the temporal.download cache, downloads were restored.
Resolved
## Summary Between July 7 and July 8, 2026, customer accounts hosted in our Prod 3 environment experienced intermittent slowness and failures when loading pipelines and resolving templates. The underlying cause was a sharp, sustained increase in internal feature-flag lookup traffic from our template-processing service to our core platform backend service. The issue presented across two separate days. On Day 1 \(July 7\), our team restored service through infrastructure level mitigations by restarting affected services and adding capacity. Because that underlying cause was still present, the same failure mode recurred on Day 2 \(July 8\). Deeper investigation on Day 2 identified the specific feature flag and code path responsible, and engineering shipped a fix so that flag check is now served from local cache. This fix was deployed across all production clusters and fully resolved the issue; it has not recurred since We are auditing all other feature flags on the platform for the same caching gap, and are implementing process and monitoring improvements described in the Preventive Measures section below to catch this class of issue earlier and reduce the risk of recurrence. ## Impact **Symptoms:** Intermittent slowness or failure loading pipelines; slow or failed template validation and resolution; intermittent connectivity/timeout errors between platform services. **Functional impact:** Pipeline executions and pipeline-studio operations that depend on template resolution were delayed or failed during the active windows described in the timeline above. Some executions required manual retry ## Root cause Our core backend service could not keep up with the request arrival rate because of a sustained, uncached feature-flag lookup driving a large and growing volume of internal traffic following a recent release. ## Mitigation * Immediate mitigation \(Day 1 and Day 2\): removed the saturated backend instance\(s\) from service and added capacity to relieve the acute bottleneck. * Permanent fix \(Day 2\): updated the template-processing service so the responsible feature-flag check reads from local cache instead of calling the backend service on every lookup, eliminating the excess internal traffic at its source. * The fix was deployed to all production clusters and verified through a return to normal internal traffic volumes and response times. ## Preventive Measures & Next Steps * Feature-flag caching audit: We are reviewing all feature flags used in high-frequency code paths across the platform to identify and close any similar caching gaps before they can cause a repeat of this issue. * Release validation: We are strengthening pre-release load testing for new feature-flag-gated code paths so that call-volume regressions of this kind are caught before reaching production. * Capacity alerting: We are adding proactive alerting on backend request-queue saturation so that this class of bottleneck is caught and mitigated automatically, before it affects customer-facing response times.
Resolved