AI incidents: ChatGPT Work disruption, Claude and ElevenLabs recoveries — 7 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

New ChatGPT Work threads remained affected in desktop version 26.1002.51308 at the 11:58 Istanbul cutoff on 7 October. The other new official notices below describe recoveries. This edition also follows up on OpenAI’s September 29 event with a newly observed postmortem and corrects an ElevenLabs coverage statement in yesterday’s incident edition.

The scan covers 4 October at 12:13 through 7 October at 11:58 Europe/Istanbul (UTC+3), overlapping the previous scan to catch updates. Previously reported events are deduplicated. All table and timeline times below are Istanbul time. Acknowledgement and resolution-post times are not automatically the actual impact boundaries. Product releases are in the separate 7 October AI news edition.

ChatGPT Work: new desktop threads still affected

OpenAI acknowledged the issue on 7 October at 05:00, then announced rollback work at 08:54. Its component list names Codex in ChatGPT Desktop. The notice gives no actual impact start or technical cause, and no resolution was posted by cutoff. Original desktop incident.

Compare your installed version with the affected version in the table and follow the provider’s recovery updates.

Claude: usage reports recovered after 100 minutes

Anthropic says errors affected the platform.claude.com usage page and Admin API usage-report requests from 03:00 to 04:40 on 7 October. The Messages API was unaffected. This explicit interval supports a 100-minute duration for usage reporting, not a 100-minute Claude chat outage. Investigation was posted at 04:20 and resolution at 05:33, after the stated recovery. No technical root cause appears in the notice. Official usage-report incident.

A separate Claude Opus 5.5 incident on 6 October concerned elevated request errors across the listed claude.ai, API, Claude Code and Cowork surfaces. Anthropic posted investigation at 15:24, identification at 15:37 and resolution at 15:43. Those posts do not establish a 19-minute impact duration or a failure of every Claude model. Official Opus 5.5 incident.

OpenAI: recovered agent, PDF and ChatGPT Go services

These are four distinct provider-confirmed events, with no published common cause. Times describe official updates unless otherwise stated; exact impact durations remain unknown.

• Workspace Agents: OpenAI said some users might not see agent responses. In Istanbul time, investigation was 7 October at 01:52, identification 04:52, mitigation/monitoring 05:45, and recovery 06:00. The original investigation was dated October 6 in UTC; this is the same event across the timezone boundary. Workspace Agents timeline.
• Computer Use: the notice lists the Agents API component, with investigation 6 October at 20:15 and recovery 23:00. It does not establish that all API products failed. Computer Use timeline.
• Large PDF processing in the Responses API: investigation was posted 6 October at 18:23, mitigation work at 18:29, and recovery at 20:50. The notice concerns that file-processing workflow, not every Responses request. PDF incident.
• ChatGPT Go conversations: investigation was 6 October at 16:35, mitigation/monitoring 16:53, and recovery 17:40. The named plan is Go; the evidence does not establish the same impact for all subscription tiers. ChatGPT Go incident.

ElevenLabs: EU SIP calling and text-to-speech recovered

EU SIP calling: a documented 2-hour-13-minute impact window

ElevenLabs’ postmortem gives an actual impact window of 12:53–15:06 on 6 October. It says roughly 66% of inbound calls and 100% of outbound calls using its EU SIP servers with static IPs crashed or timed out. Those are the provider’s figures for that specific infrastructure, not measurements of all ElevenLabs calls. Official postmortem.

ElevenLabs attributes the failure to port exhaustion: a call-teardown bug failed to release ports, accumulating over weeks. It describes periodic server restarts and enhanced port monitoring as mitigations while a permanent patch was still being investigated. The incident timeline shows acknowledgement at 13:28, monitoring at 15:14, and a resolution post at 15:51. The final post came after the stated impact ended.

Text-to-speech latency: resolved, with a correction to yesterday’s coverage

A separate notice describes elevated latency for a small percentage of ElevenLabs API text-to-speech requests. Investigation was posted 6 October at 12:09, mitigation/monitoring at 14:38, and resolution at 15:14. No exact impact start or technical cause is given. Original TTS incident.

Correction: yesterday’s edition said ElevenLabs’ history listed no October incidents. The TTS notice is timestamped four minutes before that edition’s 12:13 cutoff, so that coverage statement was incomplete. We have corrected it in place, preserving the original publication date and recording today’s correction.

OpenAI postmortem: follow-up to the September 29 disruption

This is an update to an older event, not a new October 7 outage. OpenAI’s newly observed postmortem attributes the disruption to a feature rollout whose repeated background checks exhausted shared network connections.

The provider now places impact at approximately 29 September, 20:30 Istanbul, with most impact mitigated by 21:00 and residual degradation ending around 30 September, 02:07. These correspond to September 29 at 10:30, 11:00 and 16:07 PDT. The postmortem’s own publication timestamp is not shown. We updated yesterday’s earlier coverage to link the available write-up; its October 6 status update remains an update to the September event.

What affected users can check

Match the product, component, app version and timestamp to the source notice. A provider’s resolution does not establish that an interrupted generation, task or account recovered individually. Our practical suggestion is to inspect the existing job or conversation before starting another attempt, and retain request IDs and error messages for support without sharing credentials.

For the desktop Work issue, follow the official rollback/recovery updates. For ElevenLabs telephony, check whether the call used the affected EU static-IP SIP route before associating it with that incident. Neither provider notice promises a refund or authorizes automatic replay of paid work.

Customer-review follow-up and coverage limits

We rechecked music, video and remix/stem reviews, including our existing Suno vocals, Runway retries and credits, Moises lyrics and Kits download-minutes reports. The reviewed material did not establish a new resolution. Additional quality and credit complaints remain research leads or repeat existing themes; they do not establish a common technical incident or justify another allegation-based item today.

Google coverage remains incomplete. Its Cloud product history yielded no qualifying new event in the reviewed window, but the separate Gemini API status page was inaccessible. Cloud history does not establish availability for consumer Gemini, Flow, Veo or MusicFX. Several other providers exposed only current state or unreadable historical interfaces. This is a selective daily scan, not a complete incident census or continuous monitoring.

Update timeline

• 7 October 2026, 11:58 Istanbul / 08:58 UTC: Initial edition adds nine official October 6–7 incident notices, one older-event postmortem follow-up, and the ElevenLabs correction to yesterday’s edition. ChatGPT Work desktop remained identified at cutoff.

Sources

• Anthropic: usage reporting and Opus 5.5 errors.
• OpenAI: Work desktop, Workspace Agents, Computer Use, large PDFs, and ChatGPT Go.
• ElevenLabs: EU SIP incident, its postmortem, and TTS latency.
• OpenAI: September 29 event and postmortem.
• Coverage: Google Cloud history; customer-review originals and support guidance remain linked in the four standalone reports above.

Official notices establish the events their operators acknowledge; they are not independent measurements. Original pages were checked on 7 October 2026. Unknown impact boundaries and causes remain unknown.