AI incidents: Runway upscaler failure and Claude recoveries — 24 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.

Recovery update — 25 October 11:56 Istanbul: Runway says Magnific upscaling was working again at 15:21 on 24 October and marked the incident resolved at 16:29. The original cutoff and timeline below are retained. Provider record, today’s follow-up.

Incident summary and status at original publication

Runway is investigating failed Magnific upscaling. Anthropic reports recovery from three separate model-error incidents, and OpenAI has resolved failures in ChatGPT app connections and Help Center chat. YepAPI separately acknowledges and repairs problems in its own video integration. ChatGPT web-dictation complaints are corroborated by independent accounts, with one participant subsequently reporting recovery.

This edition covers 21 October at 11:57 through 24 October at 12:29 Europe/Istanbul (UTC+3). All times below are Istanbul time. Claude’s stated impact windows are distinct from its later resolution notices; OpenAI’s intervals measure the public notice timeline because separate customer-impact boundaries are unknown.

Runway: Magnific upscalers remain unavailable

Runway says Magnific upscaling is failing and it is working with its provider to restore service. The notice was posted at 12:15 Istanbul on 24 October and remained in Investigating status at our cutoff. Its text says the cause has been identified, without explaining that cause or giving a recovery estimate. Original incident, timestamped incident record.

The record associates App, Backend and Public API components, while their aggregate indicators remain operational. The acknowledged scope is Magnific upscaling, not every Runway generation service. A separate actual impact-start time and workaround are not supplied; the notice timestamp is not sufficient to calculate the full outage duration. No recovery had been confirmed at the original cutoff. Subsequent update: Runway reported a fix and working upscaling at 15:21 on 24 October, then resolved the incident at 16:29; the actual impact-start time and technical cause remain unknown.

Claude: three distinct error incidents recovered overnight

Anthropic gives explicit impact windows of 31, 11 and 35 minutes for three separate events. Each record lists claude.ai, the Claude API, Claude Code and Claude Cowork. That component list does not mean every model failed continuously across those products.

Several models, including Opus fast mode: 02:28–02:59

The first incident’s title names Fable 5.1, Sonnet 5.5 and Haiku 5.5; its identification update also includes Opus 5.5 fast-mode requests. Investigation began at 02:37, identification followed at 02:39, and Anthropic reported improving success rates at 02:54. The later resolution notice, at 03:11, specifies the 02:28–02:59 impact window. Monitoring therefore began before the provider’s stated impact end. Original incident and full updates.

Opus 5.5: 03:23–03:34

A separate Opus 5.5 event affected requests for 11 minutes, according to Anthropic’s resolution text. Investigation was posted at 03:31 and resolution at 03:37. There is no published cause, regional breakdown or workaround in the notice. Original Opus incident.

Fable 5 and Fable 5.1: 08:00–08:35

The third record names Fable 5 and Fable 5.1 and gives a 35-minute impact window. Investigation appeared at 08:16, identification at 08:24, another identified update at 08:40, monitoring at 08:41 and resolution at 08:53. Some update text refers generically to other Claude models without identifying additional versions; we do not extend the named scope to the whole model catalog. Original Fable incident.

Anthropic says causes were identified in the first and third events but does not disclose them. No shared cause is established. Provider recovery also does not establish that a previously failed request or generation completed automatically.

OpenAI: new and refreshed ChatGPT app connections recover

Some ChatGPT app connections failed during new connections or refreshes; OpenAI marked the CIMD incident resolved at 02:43 on 24 October. At 02:17, the provider said the affected authentication metadata endpoint was available again. Original app-connections notice.

Investigation began at 01:52, and mitigation entered monitoring at 02:10. The acknowledgement-to-resolution interval is approximately 50 minutes; the notice does not establish a separate start for every user’s impact. It gives no underlying root cause, region-specific scope or user workaround. This is a connection-authentication problem, not evidence that all ChatGPT conversations or all connected-app actions failed.

OpenAI: Help Center chat restored after a separate disruption

OpenAI reports that Help Center chat recovered at 01:02 on 24 October. Investigation was posted at 00:30 and monitoring at 00:52, an approximately 33-minute notice interval. During the disruption, OpenAI directed users to support@openai.com. Original Help Center notice.

The notice does not disclose a cause or mark individual product components. It is distinct from the support-response delays reported in the previous edition. Restoring chat access does not prove every pending support request has been answered.

YepAPI: Veo integration and large-video downloads repaired

YepAPI says Veo generation works again after it changed its routing, and completed videos over 10 MB now download correctly. Its 23 October notice attributes the generation interruption to the retirement of Google’s Veo 3.1 preview endpoints on 22 October. The same notice says some completed downloads had returned empty responses. Exact failure and recovery times are not supplied, so no duration can be calculated. YepAPI’s original repair notice.

This is an acknowledged YepAPI integration incident, not a Google-wide outage. Google independently confirms the preview shutdown; generally available Veo models on another platform are a separate route. Google’s retirement notice. Related catalog and pricing changes appear in today’s product-news roundup.

ChatGPT web dictation: corroborated complaints, one reported recovery

Independent users reported delayed or missing text and poorer punctuation after a change to the web message composer’s microphone dictation. The original account was posted on 21 October at 23:49 and identifies Chrome on Windows 11; an independent account at 04:04 on 22 October reports matching behavior. This concerns speech-to-text in the composer, not a confirmed Voice Mode outage. Original account, independent account.

At 20:28 on 23 October, that second participant said browser reloads again showed the earlier dictation behavior. A different participant reported dictation working well in the Mac desktop app. These are useful recovery and counterexamples, but do not establish a general rollback or a fix for every account. Recovery, Mac counterexample.

No provider confirmation, technical cause or reliable population-wide duration was established. The thread’s speculation about processing locations or commercial motives is not evidence.

Follow-up checks and coverage limits

Google’s London record remains preliminary, without a newly verified final analysis since our earlier follow-up. ElevenLabs’ readable history contains no newly verified substantive incident for this edition; earlier promised analyses remain unverified. Google record, ElevenLabs history.

Older ChatGPT account-capacity, Work-loading and scheduled-write threads were rechecked without establishing a new general recovery. Restricted review feeds and incomplete xAI, DeepSeek, Perplexity and consumer-Gemini histories remain coverage gaps, not evidence that those services had no problems. Isolated billing, credit and quality allegations remain unpublished leads.

Sources

• Runway Magnific upscaler incident.
• Anthropic incident records: multiple models, Opus, Fable.
• OpenAI incident records: app connections, Help Center chat.
• YepAPI repair notice and ChatGPT dictation discussion.

Public community accounts are not necessarily verified customer reviews. MUVO publishes this reporting and offers competing creative tools; see our reporting method.

Update timeline

• 24 October 2026, 12:29 Istanbul / 09:29 UTC: Research cutoff. Published the ongoing Runway upscaler incident, five distinct resolved provider events, qualified YepAPI recovery and corroborated dictation reports. UTC source timestamps are retained in the research ledger. Earlier article dates and content were preserved.

• 25 October 2026, 11:56 Istanbul / 08:56 UTC: Added the provider-confirmed Runway recovery, which occurred after this edition’s original cutoff. Original publication date, URL and historical status table preserved.