AI incidents: OpenAI recovery and Google Cloud London update — 21 October 2026

Publisher disclosure: MUVO offers competing music, remix, and video tools. This report summarizes public sources and provider documentation; it is not an independent product test or a claim that MUVO is free of similar issues. Read our reporting method.

Incident summary and status at publication

OpenAI reports recovery from an overnight disruption, while Google’s London update retains a warning about a small number of customers. Separate ChatGPT Work loading complaints continue; the broader recovery notice does not resolve those individual cases.

This edition covers 18 October at 11:52 through 21 October at 12:01 Europe/Istanbul (UTC+3). It includes newly verified follow-ups to earlier incidents. It does not cover events later today.

OpenAI: ChatGPT, API and Codex degradation resolved

OpenAI marked the incident resolved at 06:28 Istanbul on 21 October. The notice lists API components including Responses, Chat Completions, Audio, Realtime, Images and Files; ChatGPT components including Conversations, Login, Work, Voice mode and File uploads; and Codex Web, API and CLI. Official incident and complete component list.

The first investigating notice was at 05:26:28, followed by mitigation and monitoring at 05:48. The acknowledgement-to-resolution interval was approximately 1 hour 2 minutes. Customer-impact start is unknown, so this is not a measured duration for every user.

OpenAI says affected services recovered. It supplies no root cause or specific workaround, and does not say failed work was replayed. Account-specific failures still need separate investigation.

Google Cloud: London recovery retains a residual-impact caveat

Google says the London disruption was mitigated for most customers, while a very small number were still believed affected. Its final notice appeared at 17:03 on 20 October, with a 17:05 modification. This updates the ongoing incident in our 20 October edition. Official notice, timestamped incident data.

Google attributed the disruption to a power event in europe-west2-b. Most workloads were recovering by 15:10. Its recorded event interval runs from 09:08 to 17:00 on 20 October: 7 hours 52 minutes. The residual-impact statement means that endpoint is not proof every customer recovered then.

Affected customers were directed to Google Cloud Support; an analysis was promised. Gemini, Vertex AI and Veo are not named in the affected-product list. This remains a regional infrastructure incident, not evidence of a consumer Gemini outage.

ChatGPT Work: loading complaints continue after the earlier support finding

Two independent accounts posted on 21 October describe continuing failures to load Work conversations. One reports recurring failures over weeks; another says the problem returns after one or two tasks and affects multiple devices. These are customer reports, not a newly acknowledged service-wide incident. First dated account, second dated account.

OpenAI Support’s earlier finding linked one supplied conversation to a rate limit and said there was no tested workaround. That finding cannot establish the cause of every new report. The second account was posted after the broader incident’s resolution timestamp, but temporal overlap alone does not prove either a shared cause or a recurrence. Support reply.

We have not reproduced the failures, and the sources do not establish a continuous outage duration, affected-user count or general recovery.

Sources

Official notices establish acknowledged events within their stated scope. Customer reports need independent corroboration and remain labeled as reports. We also rechecked creative-tool reviews and previously published issues; isolated billing and output-quality complaints did not establish additional publishable incidents.

• OpenAI: service-degradation timeline and incident history.
• Google: London incident and original public incident feed.
• ChatGPT Work: 21 October account one, account two and earlier support finding.

Inaccessible sources and incomplete histories are research gaps, not proof that a service was healthy. See the reporting method and the separate 21 October product-news edition.

Update timeline

• 21 October 2026, 12:01 Istanbul / 09:01 UTC: Research cutoff. Added OpenAI’s recovery, Google’s qualified London recovery and continued Work reports. Added a dated recovery note to the 20 October incident article while retaining its original publication date and historical cutoff.