AI incidents: OpenAI review errors, Ads recovery and scheduled-task fixes — 27 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 current status

OpenAI is monitoring a mitigation for failures in hosted Code Reviews and Security Reviews on private GitHub repositories. A separate Ads API incident is resolved. New support and customer updates also establish recovery for some previously blocked ChatGPT scheduled actions, while a SharePoint file-transfer check remains under investigation.

Research covers 24 October at 12:00 through 27 October at 12:24 Europe/Istanbul (UTC+3). The table distinguishes official incident notices from support-thread follow-ups. Notice times are not necessarily the beginning or end of customer impact.

OpenAI monitors private-repository Code Review and Security Review errors

The official notice describes elevated errors for most hosted Code Reviews and Security Reviews on private GitHub repositories. OpenAI acknowledged the issue at 11:00 Istanbul on 27 October and said a mitigation was applied at 11:21. At this edition’s cutoff the notice remains monitoring, rather than resolved. Original incident notice.

The notice does not specify a root cause, affected regions or a confirmed impact-start time. Its component list is empty, so the scope here follows the explicitly named hosted-review workflows. It does not establish that all Codex operations or GitHub itself are unavailable. No workaround is documented in the notice; check the linked status record before depending on a new hosted review.

OpenAI resolves Ads API POST-request errors

OpenAI says the affected Ads API services recovered on 26 October at 21:30 Istanbul. The first notice at 21:19 identified elevated errors on POST /ads/*; mitigation and monitoring followed at 21:25. The affected component is Ads API within Ads Platform. Official notice and timeline.

The roughly 11-minute acknowledgement-to-resolution interval measures the public notice sequence, not a proven outage duration. Actual impact start and root cause are unspecified. This is a specific advertiser API disruption, not evidence that ordinary ChatGPT conversations failed. Provider recovery also does not confirm whether a particular failed write completed; check its resulting state before repeating it.

ChatGPT scheduled actions: Gmail and Sheets recover, SharePoint has a separate remaining check

OpenAI Support says it shipped a change to scheduled-action approvals last week, and a customer independently confirms Gmail sends and Sheets updates are running again. The support post at 21:33 Istanbul on 26 October cites successful runs from Friday, 23 October. At 00:56 on 27 October, a customer says they verified the resulting emails in Sent and the changes in the spreadsheet. The exact deployment and account-recovery times are unknown. Provider update, customer confirmation.

SharePoint copying improves; file-reference transfers remain unresolved

A different participant reports verified scheduled SharePoint copies on 21, 23, 24 and 26 October, but failed file transfers on 26 October at approximately 19:45 and 20:48 Istanbul. Those test times were supplied as Pacific Daylight Time and converted to Istanbul. The reported error is BLOCKED_FILE_REFERENCE; the same author also says interactive document replacement succeeds while scheduled replacement fails. Detailed account, same-author clarification.

At 02:43 Istanbul on 27 October, OpenAI Support explicitly distinguishes this error from last week’s approval change: it comes from a separate ChatGPT check before sending a file to SharePoint. Support is checking whether scheduled tasks support that creation/replacement workflow. This is not a diagnosis of a Microsoft SharePoint outage or a confirmed general recovery. Support clarification.

Practical next step: verify the actual destination email, spreadsheet or document after a scheduled run. If a later run remains blocked, support asks for its time and timezone, with task details sent privately through the Help Center. We added dated recovery notes to the 19 October report and 23 October report; their original dates and historical cutoffs remain intact.

Conversation-loading complaints continue in an existing ChatGPT thread

Two independent participants added reports of the same conversation-loading problem on 27 October, at 10:42 and 11:42 Istanbul. One identifies a Business subscription; the other says the symptom began that day. These are customer-reported continuations of an existing issue, not a newly confirmed platform-wide outage. Neither post supplies a verified cause or recovery time. First account, second account.

Follow-up checks and evidence limits

Anthropic’s recent model-error notices and Runway’s Magnific incident remain resolved in their official histories, with no fresh material change verified. Google’s London report remains preliminary; ElevenLabs’ older promised incident write-ups remain incomplete in the accessible sources. Newly encountered isolated credit, output-quality and account complaints are retained as unpublished leads rather than combined into an invented common outage.

The scan includes music, video, voice and remix tools alongside major AI services. Some specialist feeds, review platforms and tenant-specific histories were inaccessible or incomplete. A healthy current indicator alone was not treated as proof that no earlier disruption occurred. Public community participants are not necessarily verified customer reviewers, and no account-specific failure was reproduced by MUVO.

Sources

• Official OpenAI notices: hosted reviews, Ads API.
• Original scheduled-action thread: provider update, Gmail/Sheets verification, SharePoint account, provider clarification.
• Original conversation-loading accounts: post 32, post 33.
• Follow-up histories: Anthropic, Runway, Google Cloud, ElevenLabs.

MUVO publishes this reporting and offers competing creative tools. Read our reporting method and today’s separate product-news roundup.

Update timeline

• 27 October 2026, 12:24 Istanbul / 09:24 UTC: Research cutoff. Added the hosted-review notice, Ads API recovery, scheduled-action follow-up and conversation-loading continuation. Added dated recovery notes to the 19 and 23 October reports.