October 2026 AI Tools Outage & Uptime Report
We continuously monitor 18 major AI tools including ChatGPT, Claude and Gemini. A recap of incidents recorded on providers’ official status pages in October 2026.
Longest outage: Anthropic（148h 35m）
By provider
| Provider | Incidents | Downtime | Worst state | Uptime |
|---|---|---|---|---|
| Anthropic | 11 | 155h 14m | Major outage | 79.14% |
| GitHub | 10 | 18h 52m | Major outage | 97.46% |
| Cursor | 7 | 6h 52m | Partial outage | 99.08% |
| OpenAI | 6 | 19h 18m | Partial outage | 97.41% |
| ElevenLabs | 4 | 9h 35m | Partial outage | 98.71% |
| Runway | 2 | 2h 51m | Major outage | 99.62% |
| Groq | 1 | 18m | Partial outage | 99.96% |
Incident counts and downtime cover only "major-or-worse and resolved" incidents recorded on official status pages (minor latency and unresolved ongoing items are excluded). Uptime is our own estimate based on that, not official vendor figures. Aggregated by calendar month in Japan time (JST).
Highlights of the month
October 2026 saw 41 major incidents totalling 12,780 minutes across 7 providers. Anthropic logged the most (11), followed by GitHub (10), Cursor (7) and OpenAI (6). ElevenLabs had 4, Runway 2 and Groq 1.
About 70% of the total (8,915 minutes) came from a single Anthropic incident: “Elevated errors on platform.claude.com”. It stayed open on the official status page for about six days, from October 7 22:25 to October 14 03:00 (JST). Excluding it, the month totals 3,865 minutes — below September on the same basis (4,204) and August (4,153).
Eight of GitHub’s ten incidents were at the highest severity. The longest were an Actions incident from October 6 04:11 (about 3h38m), “several services degraded” from October 7 04:57 (about 3h43m), and delays for pull requests and webhook deliveries from October 21 23:52 (about 4h30m).
At OpenAI, from October 7 11:00 to 20:53 (about 9h54m) users of the latest desktop app could not create new ChatGPT Work threads. On October 16 from 01:29 there was “widespread feature degradation” for about 2h27m.
Seven of Anthropic’s 11 incidents were elevated errors for Claude Sonnet 5.5, Fable 5.1 and other models between October 24 and 28. At 08:37 on the 24th it announced a highest-severity error spike across three models (Fable 5.1, Sonnet 5.5, Haiku 5.5); about a minute later Cursor declared “Grok models unavailable” (about 41 minutes), and about three minutes after that GitHub posted “Grok 4.6 and Grok 4.7 service issues” (minor). All three began within a few minutes, but the official notices do not say whether they shared a cause.
ElevenLabs had 4 incidents totalling 575 minutes: increased dubbing latency and failures for about six hours from October 9 01:09, and two SIP-call disruptions — EU agents from October 6 19:28 (about 2h23m) and another of about 44 minutes from October 9 00:00. Cursor had 7 incidents totalling 412 minutes, mostly around Cloud Agents and Grok Bot (for example about 2h39m from October 7 04:27).
Runway has an entry from October 18 03:55 lasting about 2h24m that the vendor titled “Scheduled Maintenance” yet recorded at the highest severity. Planned work is counted as an incident under the vendor’s own classification.
Trends & takeaways
A single long incident made up about 70% of the month. Anthropic’s uptime reads a low 79.14%, but without that one incident it is about 99.1%. An incident with a narrow scope still weighs heavily in the numbers if the vendor keeps it listed as a partial outage for a long time.
Anthropic and GitHub had the most incidents, and GitHub had the most at the highest severity (8 of 10). Failures in Actions, pull requests and webhooks — parts of the development workflow itself — stood out.
Problems around new models and features continued: Sonnet 5.5 and Fable 5.1 at Anthropic, and Grok-related issues at Cursor and GitHub. If you build a new model into your work, it is safer to keep the ability to switch to another model for a while.
Minor degradations (latency level) numbered 85: OpenAI 32, Cursor 25 and GitHub 14 — about twice the 41 major incidents.
How to read this report
Counts and downtime include only incidents recorded on providers’ official status pages that were “partial outage or worse (effectively unusable)” and resolved, by calendar month in Japan time. Minor degradations such as latency, and ongoing items with no resolution announced, are excluded so the numbers do not look larger than reality.
No items were unresolved at the end of the month.
About 70% of this month’s total is one Anthropic incident; subtracting it gives a truer picture of the month overall.
ElevenLabs’ official status page had re-published some past incidents under new IDs. Identical entries are counted once.
Uptime is our own estimate based on the major downtime above, not an official vendor SLA figure.
Each incident’s title, severity, duration and official link are on each service’s “Past outage history” page.