AI incidents: Google Cloud London disruption, OpenAI recovery and Firefly reports — 20 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

Google Cloud has acknowledged a London-region disruption, and OpenAI has resolved a separate file-download incident. New support replies also clarify one ChatGPT Work loading case and the handling of Adobe Firefly generation complaints. These are distinct events, with different evidence and affected products.

The research window is 17 October at 11:55 through 20 October at 11:52 Europe/Istanbul (UTC+3). Statuses below reflect that cutoff, not the rest of the day.

Google Cloud: London errors remain under investigation

Google reports elevated errors and intermittent network failures in europe-west2, beginning at 09:08 Istanbul on 20 October. Its first public notice was posted at 11:35. The affected list includes Cloud Run, Compute Engine, Kubernetes Engine, Cloud Storage, Cloud SQL and other regional infrastructure services. Official incident, original incident feed.

The notice supplies no root cause or workaround. It promises another update by 12:30 Istanbul; that is an update target, not a recovery estimate. No resolution was recorded at our cutoff.

Scope matters: Gemini, Vertex AI and Veo are not named in the affected-product list we retrieved. This is a regional cloud-infrastructure incident relevant to developers hosting workflows there; it does not establish a Gemini-wide or AI-generation outage. Check the notice against your project’s region and services. We cannot calculate a completed outage duration while recovery remains unknown.

OpenAI: file-download errors recovered on 19 October

OpenAI marked its file-download incident resolved at 17:14 Istanbul on 19 October. The incident named the API’s Responses, Images and Files components, plus ChatGPT File uploads and Image Generation. OpenAI’s notice described elevated errors and subsequent full recovery. Original incident and component list.

The identified notice was timestamped 16:31:52 Istanbul, followed by resolution at 17:14:03. That is approximately 42 minutes between acknowledgement and resolution, not a proven duration of every customer’s disruption. The source does not establish a separate customer-impact start, root cause or workaround.

The recovery is OpenAI’s service-level statement; an individual failed upload, download or generation may still need checking. The notice does not say that failed work was automatically replayed or that every affected account recovered simultaneously.

ChatGPT Work: support traces one loading failure to a rate limit

OpenAI Support says its logs show that a rate limit prevented one supplied Work conversation from loading at the reported time. The reply was posted at 01:01 Istanbul on 20 October, and explicitly says there is no tested workaround yet. Support’s findings.

This is a material follow-up to the conversation-loading reports in our 18 October incident edition. It does not establish that all loading errors share that cause, that conversations have been deleted, or that the problem is fixed.

Support asks another affected reporter to add the thread link to their existing private support request so cases can be connected. A user’s claimed recovery after logging out does not make that a tested general remedy. No defensible outage duration or affected-account count is available.

Adobe Firefly: further Seedance generation complaints and a manager response

Two independent accounts dated 17 and 19 October report continued generation failures in Adobe’s Seedance 2.0 bug thread. One describes repeated retry errors despite changing browsers and accounts; the other says generation fails when started. These are public user reports, not independently reproduced tests. Dated accounts and provider response.

On 19 October, an Adobe Community Manager said the thread had been shared with the team and would be closed to further replies. Affected users were asked to create a new post with details and examples so individual cases can be investigated.

That response confirms escalation, not a diagnosed outage or recovery. The thread also contains different symptoms and older model-policy advice, which do not establish one common cause. Exact plan, region and model selections are not supplied for every reply.

This finding concerns Adobe’s hosted workflow. It is not evidence that ByteDance’s upstream Seedance service failed everywhere. The source provides dates without a timezone or precise event boundaries, so we do not assign a duration.

Sources

We inspected original status notices, public incident data and dated community replies. Provider acknowledgements establish only their stated scope; community reports do not establish prevalence or verified purchases. We have not reproduced these failures in private accounts.

Music, video and remix/stem reviews were also checked. Isolated billing allegations, vague quality complaints and incomplete source dates remain research leads. Inaccessible histories are coverage gaps, not evidence of healthy service.

• Google: London incident and original JSON history, including the provider’s impact-start timestamp.
• OpenAI: file-download recovery and Work support reply.
• Adobe: Seedance bug thread, dated accounts and manager reply.

Product launches and current AI discussions are covered separately in the 20 October news edition. See our reporting method.

Update timeline

• 20 October 2026, 11:52 Istanbul / 08:52 UTC: Research cutoff. Added Google’s ongoing London disruption, OpenAI’s file-download recovery, the Work support finding and Adobe community follow-up. Earlier editions retain their original content and dates.