AI access, coding agents & developer services · Daily incidents

AI incidents: ChatGPT Work, Grok Bot and Composer recoveries — 6 November 2026

OpenAI resolves ChatGPT Work, Dots and Codex Web errors. Grok Bot, Composer 2.5, Fireworks and GitHub report separate recoveries, with scope and timelines.

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 the cutoff

Update, 7 November: Fireworks announced restoration of its older affected deployment, training and batch-inference services at 21:28 Istanbul on 6 November and moved the incident to monitoring. Provider recovery notice. The original cutoff account below is retained; see the 7 November follow-up.

OpenAI, Grok Bot, Cursor’s Composer 2.5 and Fireworks have reported recoveries from separate disruptions. GitHub also resolved two developer-service incidents. These notices describe different products and affected components; they do not establish a simultaneous outage across AI providers.

This edition covers 3 November at 11:58 through 6 November at 12:15 Europe/Istanbul (UTC+3), including a 48-hour overlap with the previous successful scan. Previously covered events are deduplicated. Times below are Istanbul time, and recovery statements are the providers’ assessments rather than confirmation that every account or failed job recovered.

ChatGPT Work and Dots: degraded performance ended before the incident closed

OpenAI marked the incident resolved at 07:31 on 6 November, after mitigation and monitoring began at 20:05 on 5 November. Its affected-component list includes ChatGPT Work, Dots and Codex Web. OpenAI incident notice.

Milestone Istanbul time
First investigation notice 5 Nov 18:25
Elevated errors identified 5 Nov 19:57
Mitigation applied; monitoring 5 Nov 20:05
Resolution notice 6 Nov 07:31

OpenAI’s component timeline records degraded performance for Work and Dots from 18:25:03 to 20:05:56 on 5 November, about 1 hour 41 minutes. That measures the provider-recorded component interval, not the earliest possible user impact. It is not a 13-hour continuous outage: the incident remained open during monitoring. Codex Web has a shorter recorded interval, 19:26:31–19:57:41.

The notice does not identify a root cause, an affected region, a model-specific failure or a user workaround. It does not establish failure of all ChatGPT conversations or every OpenAI API.

Grok Bot and Composer 2.5: two distinct resolved events

Grok Bot errors were resolved on 5 November; a separate Composer 2.5 request issue was resolved later that day. The official status feed lists both events separately. Cursor’s status address now redirects to the status site at status.x.ai, whose component labels distinguish these products. Original incident feed.

For Composer, those surfaces are Automations, Bugbot, Cloud Agents, CLI, Cursor IDE on desktop and Cursor IDE on iOS. The notice does not say every feature on those surfaces failed. Composer notice.

The Grok Bot event predates yesterday’s cutoff but was newly verified in this scan. Neither notice supplies an impact start distinct from acknowledgment, a root cause or a workaround, so we do not calculate outage durations. Grok Bot notice.

Fireworks: Muse Spark Open US recovers; older deployment issue remains open

Fireworks reported a brief degradation of its hosted Muse Spark Open US service, then restored it. Investigation began at 01:36 on 6 November and resolution followed at 01:40. The recorded component interval is 01:36:24–01:40:32, about four minutes. Symptoms are described only as impaired performance; cause and workaround are not supplied. This is a Fireworks-hosted service event, not evidence of an outage in Meta’s consumer app. Fireworks incident.

The separate deployment creation, dedicated training and batch inference degradation first acknowledged on 30 October at 23:51 Istanbul remains identified, with no fresh recovery statement at this cutoff. It is an unchanged continuation of earlier coverage, not a new outage. Ongoing Fireworks notice.

GitHub: workflow starts and a later multi-service disruption recover

GitHub resolved two separate incidents that could interrupt development workflows. These are developer-infrastructure reports, not evidence that every coding assistant stopped working.

Those timestamps are acknowledgment and response milestones. The notices do not establish independent impact-start boundaries, so they are not used as outage durations.

What the rechecks establish—and what remains unknown

Additional evidence, 7 November: Two ElevenLabs EU notices now document resolved choppy audio in Eleven v4 Turbo Agent conversations on 3–4 November. Their initial public-availability time is unverified. See today’s retrospective coverage for scope and event intervals. The following paragraph records the earlier scan’s findings.

Claude’s public incident API showed no new acknowledgment or material update in this window. Google AI Studio’s rendered history listed its latest incident on 30 October; Google Cloud’s London postmortem update was already covered on 5 November. ElevenLabs’ global, EU, India and Singapore histories still listed the resolved 4 November sign-in event, with no later official incident found in those histories. These checks do not rule out unreported account-level problems. Claude history, Google AI Studio history, Google Cloud incident history, ElevenLabs histories.

We also revisited the creative-tool watchlist and accessible original reviews and community threads. Repeated complaints without a new verified recovery or material change were retained as research follow-ups. Search snippets and outage-aggregator graphs were not treated as confirmed events. Some review pages blocked direct retrieval, and several product pages exposed incomplete histories; the research ledger records those gaps.

Practical analysis: If a task failed during one of these intervals, check its actual result or job state before repeating an action that could create a duplicate output or charge. A resolved status notice is a reason to recheck your workflow, not proof that an earlier failed action was replayed successfully.

Sources

The incident links beside each section are original provider notices. Exact UTC timestamps, incident identifiers, source retrievals and coverage gaps are retained in MUVO’s research ledger. Read our reporting method for evidence standards. Product announcements are collected separately for the weekly news edition.

Update timeline