AI access, conversations, billing & developer services · Daily incidents

AI incidents: ChatGPT, ElevenLabs and Bolt recoveries — 5 November 2026

Paid ChatGPT messages, ElevenLabs sign-ins and Bolt prompting recover. GitHub and Fireworks report separate disruptions; Google revises London repair targets.

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

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 has restored affected paid ChatGPT conversations, ElevenLabs has resolved its password sign-in issue, and Bolt has closed its prompting incident. Separate GitHub and Fireworks disruptions are also resolved. Customer reports about ElevenLabs credit access and ChatGPT image consistency remain distinct from those provider-confirmed recoveries.

This edition covers 2 November at 11:54 through 5 November at 11:58 Europe/Istanbul (UTC+3), with previously reported events deduplicated and older follow-ups labeled. All event times below use Istanbul time. Resolution is the provider’s assessment, not proof that every account or failed job recovered.

Product or workflow Status at cutoff First notice → resolution, Istanbul
Paid ChatGPT conversations Resolved 5 Nov 01:38 → 03:32
ElevenLabs password sign-in Resolved; update to yesterday’s report 4 Nov 11:21 → 13:52
Bolt prompting Resolved; update to earlier coverage 3 Nov 16:26 → 5 Nov 01:50
Copilot Mission Control task listing/creation Resolved 4 Nov 11:45 → 13:01
Fireworks deployment, dedicated training and batch inference Older incident still identified Notice 30 Oct 23:51; no verified resolution

These are notice intervals, not measured end-to-end outage durations. Providers generally do not disclose a separate customer-impact start.

OpenAI: errors sending messages on paid ChatGPT plans resolved

OpenAI marked the affected Conversations component recovered at 03:32 on 5 November. Its first identified notice at 01:38 said some users on paid ChatGPT plans could encounter errors when sending messages. A mitigation was applied and monitoring began at 01:49. Original incident.

The source does not identify a technical cause, a model-specific failure, an affected region or a workaround. It does not establish an API-wide or all-plan outage. The interval from first acknowledgement to resolution is approximately 1 hour 54 minutes; the actual onset of impact is unknown.

ElevenLabs: sign-ins recovered, while separate credit-access complaints remain

Password sign-in recovery on 4 November

ElevenLabs resolved its intermittent password sign-in failure at 13:52 on 4 November. After the 11:21 investigating notice, it identified the issue at 11:48, continued work at 13:06, and released a fix at 13:21 before monitoring. Its closing update says it saw no further errors and users could sign in. Original regional incident.

The Global (United States) status page said other sign-in methods were working during the incident. It did not disclose the underlying cause or say all audio, music, video or API generation was unavailable. We added the recovery to the 4 November edition.

Customer-reported payment blocks despite remaining credits

Three independent Trustpilot accounts with experience dates of 3–4 November describe being unable to use their accounts or remaining credits after payment-related problems. One describes an accidentally selected top-up; another links the restriction to changed payment details and an unsuccessful charge; a third alleges a payment-system error. These are differing account circumstances, not proof of one shared defect. 4 November account, 3 November payment-details account, 3 November blocked-account report.

We read the dated accounts on the original review listing; individual review pages were restricted. No provider reply or verified recovery was visible for these cases. A separate 4 November reviewer reports helpful support and restored credits after audiobook-quality problems—useful counterevidence, but a different case. Support recovery account.

These allegations do not establish prevalence, unauthorized billing, fraud or a service-wide payment outage. The password sign-in fix above does not resolve them.

Bolt: prompting incident resolved after crossing two midnights

Bolt marked prompting restored at 01:50 on 5 November, closing the same incident first acknowledged at 16:26 on 3 November and identified at 21:27 that evening. Official timeline.

This is a continuation of yesterday’s active incident, not a new outage on 5 November. The provider does not supply a defensible impact-start timestamp, technical root cause or account-level recovery confirmation. We have updated the earlier report’s status.

GitHub: Mission Control and development workflows recovered on 4 November

Copilot Mission Control task errors

GitHub reported approximately 2.5% errors when listing or creating Mission Control tasks, attributing them to upstream database throttling. Its first notice was 11:45, with that scope clarified at 11:47. At 12:11, it said a specific client version had been identified and a mitigation was producing partial recovery; it closed the incident at 13:01. The client version was not named. Mission Control incident.

This event began before yesterday’s cutoff but was newly verified in this scan. It does not establish failure of every Copilot model or IDE feature.

Git, API, pull-request and Actions disruptions

Three separate GitHub notices describe additional recovered failures or delays on 4 November. They matter to development and coding-agent workflows, but should not be collapsed into one continuous AI outage.

Original incident Reported scope and response Notice → resolution, Istanbul
Git, Issues and Pull Requests Errors/intermittent failures; API also affected. Main components mitigated at 14:30; Actions at 14:43. 12:49 → 14:50
Actions, Pages and Webhooks Event-triggering issue; delayed job-status updates. Webhooks mitigated at 19:45; Actions/Pages at 19:48. 19:05 → 20:01
Later Actions incident Further event-triggering degradation; mitigation described at 20:53 and monitoring at 21:07. 20:40 → 21:16

GitHub promises detailed root-cause analyses for these events. None was verified by our cutoff; the symptom descriptions alone do not establish a common cause.

Postmortem update, 7 November: GitHub now links the Actions/Pages event and later Actions recurrence to an event-routing configuration change, confirms Copilot coding-agent delays, and records two impact periods on 4 November: 18:49–19:48 and 20:20–20:55 Istanbul. Combined summary. Our 7 November report explains the recovery and recurrence.

Fireworks: seven model-endpoint incidents resolved; older service issue remains open

Fireworks closed seven additional serverless performance incidents affecting its hosted model endpoints. Each row represents a distinct official notice. The provider does not disclose a technical cause, exact impact start or workaround for these short events.

Fireworks endpoint and original notice First notice → resolution, Istanbul
Muse Spark Open US 4 Nov 20:54 → 4 Nov 20:57
Muse Spark Open US 5 Nov 03:03 → 5 Nov 03:49
Muse Spark Open US 5 Nov 04:12 → 5 Nov 04:27
Deepseek V4.1 Flash US 5 Nov 04:23 → 5 Nov 04:26
MiniMax M3.1 Flash 5 Nov 06:35 → 5 Nov 06:40
Muse Spark Open US 5 Nov 06:37 → 5 Nov 06:47
Muse Spark Open US 5 Nov 07:15 → 5 Nov 07:18

These are Fireworks-operated endpoints, not evidence that Meta, DeepSeek or MiniMax’s own services were down. Repeated Muse Spark notices do not establish continuous downtime between them.

The separate 30 October service-degradation notice still reports problems with deployment creation, dedicated training and batch inference. It remains identified, with no newly verified resolution. Its scope differs from the resolved serverless events above.

Google Cloud: London repair targets move to 13–17 November

Google’s updated postmortem extends the planned completion dates for electrical repairs after the 20 October London outage. It now targets 13 November for repairing and restoring the damaged utility feed, and 17 November for transfer back to the normal power configuration and associated generator/telemetry validation. These replace the earlier 30–31 October targets. Updated provider report.

The report still records the PLC configuration fix as completed on 26 October. The 09:00–15:08 Istanbul main-impact interval on 20 October remains unchanged, as does the distinction between the affected London infrastructure and a Gemini-wide outage. The new repair dates are estimates, not completed remediation or evidence of a new service interruption. We added the revised targets to the original 20 October report and 28 October postmortem coverage.

ChatGPT images: continuing customer reports, with a limited workaround

Independent community accounts on 5 November continue to describe lost subject identity and difficulty following reference images. This extends the 4 November customer-reported issue; it is not a provider-confirmed new outage. Subject-identity account, reference-image account.

One participant reports that character reference sheets help with simpler images, while saying detailed scenes remain difficult. Another offers a reference-sheet workflow. Partial improvement, suggested workflow. That is a user suggestion, not a validated general fix. The sample cannot establish prevalence, a shared model version, cause or duration.

OpenAI’s new Images 2.6 announcement belongs in the news edition. A release announcement does not by itself prove these account-specific complaints have recovered or identify what caused them.

Sources

Original notices, timestamped histories and dated firsthand accounts are linked beside each finding. We also rechecked Claude history, OpenAI history, Google AI Studio status, Google Cloud’s incident feed, ElevenLabs history and the wider creative-tool watchlist. No absence of a notice is treated as proof of uninterrupted service.

Restricted review pages, incomplete specialist histories, tenant-only health dashboards and an unavailable browser session limit coverage. Community accounts are not necessarily verified purchases. We did not run paid generations or independently reproduce the failures. Check your specific product, account and job against the provider’s scope before treating a status-page recovery as a fix for it.

MUVO publishes these reports and offers competing creative tools. See our reporting method.

Update timeline