Internet & cloud outages — September 2026

No internet-wide outage is in progress as of about 20:20 UTC on September 30, 2026. AWS, Google Cloud, Azure, Azure DevOps, GitHub, Anthropic, Discord, Slack, Atlassian, Vercel, Heroku, DigitalOcean, Stripe and the smaller AI platforms we track all report normal operation. Today's open incidents are small and mostly recovering. **OpenAI: elevated API latency (recovering)** OpenAI opened an incident at 17:05 UTC for elevated latency on some API requests, with Chat Completions and Responses marked degraded. A mitigation is in place and OpenAI is monitoring recovery. A separate issue with elevated errors in ChatGPT Space Pages, open since 02:20 UTC, is also in monitoring. Source: https://status.openai.com **Cloudflare: Madrid performance and Workers build delays (monitoring)** Cloudflare reported network performance issues in Madrid from 10:31 UTC and delays to Cloudflare Workers builds from 15:33 UTC. Fixes are deployed for both and Cloudflare is monitoring. The Asia-Pacific network degradation that Cloudflare had open since September 23 no longer appears among its active incidents. Two older low-impact issues remain open: Cloudflare One clients being wrongly challenged on some sites (since September 22) and WARP geolocation errors (since August 27). Source: https://www.cloudflarestatus.com **Akamai: APAC origin errors linger** Akamai's Canada edge delivery incident from September 28 is no longer listed. Its APAC origin connectivity incident stays in monitoring while an upstream carrier finishes repairs on the damaged subsea cables. Akamai says immediate impact has subsided, but intermittent errors are possible until the work is done. Source: https://www.akamaistatus.com **Twilio and Zoom: carrier-level issues** Twilio is tracking SMS failures from some short codes to several networks in Chile (recovering) and SMS delays and failures to Metfone, Viettel, Beeline and Sotelco subscribers in Cambodia (cause identified). Zoom is still working with a provider on Polycom desk phones that went offline on September 25, which is affecting Zoom Phone devices. Sources: https://status.twilio.com, https://www.zoomstatus.com **Recently resolved: Spotify outage (September 29)** Spotify went down for a large number of users on Tuesday, September 29. Downdetector reports started climbing around 9:30 to 9:45 AM ET (13:30 to 13:45 UTC) and peaked near 10:41 AM ET. Downdetector received more than 200,000 reports from 173 countries, about 46% of them from the US. Most complaints were about the mobile app (roughly 58%) and audio streaming (21%). Service came back later that morning. Spotify has not given a public cause. This happened after our September 29 report was filed, so it appears here. Source: https://www.forbes.com/sites/zacharyfolk/2026/09/29/spotify-outage-thousands-of-users-report-problems-with-music-streaming-app/ **Other notes** Crowd-sourced trackers show scattered reports for LinkedIn and Grok today, but neither company has confirmed a problem and we could not verify real impact. Microsoft retires the Azure NVv4 GPU VM series today. That is a scheduled retirement, not an outage. Meta's status page could not be read this run, and we found no reports of a Meta disruption. **Bottom line:** the internet is up. Today's incidents are regional or limited to single products, and most are already recovering.

No widespread internet outage is under way this morning (checked around 10:30 UTC). AWS, Google Cloud, Azure, Azure DevOps, GitHub, OpenAI, Anthropic, Discord, Slack, Atlassian, Vercel, Heroku, DigitalOcean, Stripe and the AI platforms we track all report normal operation. A few regional incidents are still open. Ongoing: - Akamai: Edge Delivery issues in Canada. Opened 2026-09-28 at 16:22 UTC and still marked "investigating" at the last update (23:01 UTC). Akamai has not described the scope. Akamai also still lists origin connectivity errors in APAC, which it attributes to subsea cable repairs by an upstream carrier, and higher latency on routes between Singapore, Hong Kong and Japan during planned carrier maintenance that runs until October 3. (https://www.akamaistatus.com) - Cloudflare: "Network Performance Degradation, Asia-Pacific" has been open since September 23. Cloudflare says intermittent connectivity problems affect a small subset of customers because regional capacity is badly degraded. This matches the subsea cable work Akamai reports. (https://www.cloudflarestatus.com) - Twilio: minor carrier-specific messaging problems, including SMS delivery receipt delays to CTBC Brazil (opened September 28) and older MMS delays in Australia and to GCI in the US. (https://status.twilio.com) - Zoom: a partial degradation, open since September 25, that takes Polycom devices offline for some customers. (https://www.zoomstatus.com) Resolved in the past 24 hours: - GitHub: Copilot Code Review could not complete reviews on September 28 from about 21:16 to 22:08 UTC. GitHub rated it critical and reverted a change to fix it. Other GitHub services were not affected. (https://www.githubstatus.com) - Cloudflare: elevated error rates for traffic in Eastern North America on September 28, from about 19:50 to 20:20 UTC. Also a short incident with stuck Flux hotswap models on Replicate, whose status page now lives on Cloudflare's status site. Other recent incidents from the past 72 hours, all resolved: Cloudflare R2 problems in Western North America and a Cloudflare CASB incident (September 26). Consumer ISPs: crowd-sourced sites showed small, scattered report spikes overnight for Spectrum, Cox, Megacable and a few regional networks. Nothing points to a large outage at any of them. On September 28, user reports also spiked for Life360 and several online games (Roblox, GTA Online, Rainbow Six Siege, Escape from Tarkov), but none of these was confirmed as a major outage. Routing: no significant new BGP incident was reported today. Overall: the internet is up. Keep an eye on Akamai in Canada and on APAC connectivity while the subsea repairs continue.

No internet-wide outage is in progress this Monday morning (checked around 12:30 UTC). AWS, Google Cloud, Azure, GitHub, OpenAI, Anthropic, Discord, Vercel, Atlassian, Slack, Stripe, DigitalOcean and Heroku all report normal operation. The open issues are regional and partial, and most of them are in Asia-Pacific. **Asia-Pacific latency (ongoing)** - **Akamai** has two open incidents in APAC. The first covers origin connectivity problems and elevated errors caused by congestion after unplanned maintenance on upstream subsea cables. Akamai says the carrier expects repairs in early October. The second is higher latency at peak hours in Singapore, Hong Kong and Japan while an upstream provider does maintenance, scheduled for September 18 to October 3. Akamai says core delivery is still working. ([Akamai Status](https://www.akamaistatus.com/)) - **Cloudflare** is still working on intermittent connectivity degradation for a small subset of customers in Asia-Pacific. The incident has been open since September 23 at 18:42 UTC. Cloudflare also has routine datacenter maintenance today in Stockholm, St. Louis, Los Angeles, Brussels and Honolulu, which normally shifts traffic to nearby locations without visible impact. ([Cloudflare Status](https://www.cloudflarestatus.com/)) - Background: the Australia-Singapore Cable (ASC) has been broken on its Indonesia-Singapore segment since September 4, and Vocus has located the fault in Indonesian waters. That has pushed some Australia-to-Singapore traffic onto longer routes. We have no confirmation that the cable is linked to the Akamai incident. **Messaging and voice (ongoing, limited)** - **Twilio** is investigating voice call failures from some of its mobile numbers to Vodafone UK (since 10:45 UTC). It is also monitoring recovery from SMS delivery failures to several networks in Spain that began at 08:26 UTC. Several smaller carrier-specific delivery issues remain open, including MMS to GCI (US) and Australia and SMS to Datora (Brazil). ([Twilio Status](https://status.twilio.com/)) - **Zoom Phone** users on Polycom desk phones: an issue that knocked many devices offline has been open since September 25. Zoom said on September 27 that service is back for most affected users and that it is still working with its provider on the devices that remain offline. ([Zoom Status](https://www.zoomstatus.com/)) **Recent regional disruptions (past 72 hours)** - A nor'easter along the US Northeast coast on September 26-27 caused power and internet outages. Optimum reported about 5,000 affected internet customers in Connecticut on Saturday evening. ([Yahoo News](https://www.yahoo.com/news/weather-news/articles/thousands-connecticut-without-internet-noreaster-231905672.html)) - Telus (Canada) had a spike in user outage reports early on September 27, around 2 AM Eastern. The provider has not published any details. - Cursor's Grok Bot had two short service degradations on September 25-26. Both were resolved. **Long-running** - AWS's Middle East regions (me-central-1 UAE and me-south-1 Bahrain) are still listed as disrupted after infrastructure damage in early March. AWS's most recent update, on September 15, says some resources cannot be restored. **Bottom line:** the internet is up. If you are reaching sites hosted in or through Singapore, Hong Kong or Japan, you may see higher latency until early October.

No widespread internet outage is underway this Sunday. Every major cloud platform reports normal operations: Google Cloud, Microsoft Azure, Azure DevOps, GitHub, OpenAI, Anthropic (Claude) and Discord all show no active incidents. AWS's public health dashboard lists no new events; the only open items are the long-running Middle East region (UAE and Bahrain) notices dating back to early March. The ongoing issues are regional and minor, and most of them are in Asia-Pacific: - Akamai (APAC): Akamai is still monitoring origin-connectivity problems that it put down to "upstream congestion related to unexpected maintenance on subsea cables". A separate advisory warns of higher peak-hour latency on routes through Singapore, Hong Kong and Japan while upstream provider maintenance runs through October 3. Akamai says core services are still operational. https://www.akamaistatus.com - Cloudflare (APAC): Since September 23, Cloudflare has been mitigating intermittent connectivity problems that affect a small subset of Asia-Pacific customers. Two older, smaller issues are also still open: incorrect challenges for Cloudflare One clients (since Sept 22) and incorrect geolocation for WARP users (since Aug 27). Cloudflare has scheduled extensive datacenter maintenance for September 28-29, which may briefly reroute traffic or add latency in some regions. https://www.cloudflarestatus.com - Zoom Phone: Since September 25, some customers using Polycom desk phones have seen their devices go offline. Zoom has identified the cause. https://www.zoomstatus.com - Twilio: Several carrier-specific messaging delays are open, including MMS to GCI (US), SMS to Datora (Brazil) and MMS in Australia. None of them is a platform-wide outage. https://status.twilio.com Also checked and operational: Atlassian, Slack, Vercel, Heroku, DigitalOcean, Cursor, Groq, ElevenLabs, Cohere and Stability AI. We could not get a reliable live reading from Stripe's or Meta's status pages this run, and we have seen no reports of problems with either. Past 72 hours: GitHub had a multi-service degradation on September 23-24, which hit Projects, API requests and organization creation and was traced to database replicas, plus a billing-update issue on September 24. Both are resolved. On September 26, crowd-sourced reports showed short-lived spikes for Peacock, AccuWeather and Canara Bank. One third-party tracker flagged a connectivity drop in El Salvador today. IODA's own alert feed shows no alert for the country in the past 48 hours, so we are treating that report as unconfirmed. Bottom line: the internet is up. Expect somewhat higher latency on some Asia-Pacific routes until the subsea cable maintenance ends.

No widespread internet outage is under way this Saturday. A handful of regional and product-specific incidents are open, and one short AI-platform outage from Friday evening has been resolved. **Ongoing** - **Cloudflare — Asia-Pacific network performance (day 3):** Cloudflare opened an incident on September 23 for network performance degradation in Asia-Pacific. Its latest update (Sept 25, 08:50 UTC) says it continues to mitigate intermittent connectivity problems for a small subset of customers, caused by significant loss of capacity in the region. Cloudflare has not named a root cause. Akamai also has two open APAC items from September 22: origin connectivity errors (now monitoring) and higher latency during upstream-provider maintenance in Singapore, Hong Kong and Japan. Recent subsea cable faults off Western Australia may be related, but no provider has confirmed that. [Cloudflare Status](https://www.cloudflarestatus.com/) · [Akamai Status](https://www.akamaistatus.com/) - **Zoom Phone — Polycom devices:** Since Sept 25 (~17:45 UTC), Zoom has been working an issue where some customer Polycom desk phones go offline. The issue has been identified; Zoom Phone Global is shown as degraded. [Zoom Status](https://www.zoomstatus.com/) - **Twilio:** Several carrier-specific messaging and voice issues remain open (MMS delays to GCI US, SMS delays to Datora Brazil and MTN Cameroon, voice to Hong Kong). These are routine carrier-level items, not a platform outage. [Twilio Status](https://status.twilio.com/) **Recently resolved** - **OpenAI — ChatGPT and Codex (Sept 25):** Starting around 22:55 UTC Friday (3:55 p.m. PT), users reported errors with ChatGPT and Codex, and Downdetector reports peaked near 8,000. OpenAI marked a critical Codex incident from 22:58 to 23:54 UTC, with API-key login offered as a workaround. The service recovered within about an hour. [OpenAI Status](https://status.openai.com/) · [GV Wire](https://gvwire.com/2026/09/25/chatgpt-down-for-thousands-of-users-friday-downdetector-reports/) **Long-running** - **AWS Middle East:** The ME-CENTRAL-1 (UAE) and ME-SOUTH-1 (Bahrain) disruptions that began in March are still open. AWS's September 15 update said some data hosted only in the affected zones cannot be recovered. [AWS Health](https://health.aws.amazon.com/health/status) **All clear** AWS (outside the Middle East), Google Cloud, Azure, Azure DevOps, GitHub, OpenAI (now), Anthropic, Discord, Atlassian, Slack, Vercel, Heroku, DigitalOcean, Stripe, Cursor, Groq, ElevenLabs, Cohere and Stability AI all report normal operation. No major BGP or DNS events were found. Crowd-sourced reports show only scattered, service-specific complaints (e.g., AccuWeather, NTT, Turk Telekom), with no sign of a broad outage. Meta's status page could not be read by our automated check, so we cannot confirm its status. We found no public reports of Meta disruption.

No widespread internet outage is under way this morning (checked around 10:50 UTC). The main open issue is regional: subsea cable damage in Asia-Pacific is still limiting capacity. There is also a smaller Cloudflare Durable Objects incident, which is recovering. **Asia-Pacific: subsea cable faults (ongoing, day 5)** Cloudflare reports that since September 21 at 02:20 UTC, multiple subsea cable outages have caused congestion between its Tokyo and Singapore data centers. It has rerouted traffic, and its latest update (08:50 UTC today) says a small subset of customers still sees intermittent connectivity degradation while it works with vendors to restore capacity ([Cloudflare Status](https://www.cloudflarestatus.com)). Akamai has a related open incident for increased latency in Singapore, Hong Kong and Japan during upstream provider maintenance, plus origin connectivity errors in APAC that it is monitoring ([Akamai Status](https://www.akamaistatus.com)). This adds to an already strained region: the Australia–Singapore Cable has been cut in Indonesian waters since September 4, which has pushed some Australia–Asia traffic onto much longer routes ([Light Reading](https://www.lightreading.com/cable-technology/asc-cable-break-disrupts-key-australia-singapore-route)). Users in the region may see slower page loads and higher latency, not outright outages. **Cloudflare Durable Objects (monitoring)** Cloudflare saw elevated errors for Durable Objects starting around 00:00 UTC. A fix went out, errors came back at about 02:30 UTC, and a second fix was deployed at 08:49 UTC. The incident is now in monitoring. Apps built on Workers that use Durable Objects may have returned errors overnight. Cloudflare also resolved two short incidents early today: delayed Secondary DNS zone transfers (DNS resolution itself was not affected) and a problem editing compression rules. **Other notes** - Twilio lists several carrier-specific SMS and voice delays, including higher latency for Verizon on its Identity Gateway API in the US. These are routine, per-carrier issues. - Replicate (now shown on Cloudflare's status page) had intermittent 500 errors yesterday from about 20:40 to 22:35 UTC after a backend component rebooted unexpectedly. That incident is resolved. - Meta's status page didn't return readable data during this check, so we couldn't confirm Facebook/Instagram/WhatsApp status there. Outage-report searches showed nothing notable. **All clear** AWS, Google Cloud, Azure, Azure DevOps, GitHub, OpenAI, Anthropic (Claude), Discord, Slack, Zoom, Atlassian, Vercel, Heroku, DigitalOcean, Stripe, Cursor, Groq, ElevenLabs, Cohere and Stability AI all report normal operations. We found no new BGP incidents today. The internet is up. Regional slowdowns continue in Asia-Pacific.

No internet-wide outage is in progress as of Thursday morning (12:00 UTC). The main story is regional: slower routes across Asia-Pacific after several subsea cable faults. Around the edges there is a handful of small, carrier-specific voice and SMS issues. **Asia-Pacific: subsea cable congestion (ongoing)** Cloudflare says that since September 21 at 02:20 UTC, "multiple subsea cable outages have caused congestion between Tokyo and Singapore datacenters." It has moved traffic onto other routes and is working with vendors to restore capacity. The incident has been marked "identified" since September 23. Several of its Asia-Pacific locations show as degraded ([Cloudflare status](https://www.cloudflarestatus.com/incidents/8wmvkv5jkf15)). Akamai reported the same problem on September 22: errors reaching origin servers in APAC, "likely due to upstream congestion related to unexpected maintenance on subsea cables." It is now monitoring. Separately, Akamai has upstream maintenance on routes linking Singapore, Hong Kong and Japan that runs until October 3, so some extra latency there may continue ([Akamai status](https://www.akamaistatus.com/)). This comes on top of the Australia–Singapore Cable (ASC) break of September 4 in Indonesian waters, which Vocus is still routing around ([DCD](https://www.datacenterdynamics.com/en/news/asc-cable-break-causes-connectivity-disruption-between-perth-and-singapore/)). Who is affected: users and services whose traffic crosses Asia-Pacific, especially Japan–Singapore paths. Expect higher latency and occasional timeouts, not a loss of connectivity. **Smaller incidents** - **Zoom**: Zoom Phone and Zoom Contact Center inbound calls to Zoom India numbers are failing for some Airtel subscribers. Outbound calls work. Under investigation ([Zoom status](https://www.zoomstatus.com/)). - **Twilio**: The usual mix of carrier-specific SMS and voice delivery problems, including MTS Belarus, Datora Brazil, MTN Cameroon, Telkomsel Indonesia, and Vodafone UK and Norway voice routes ([Twilio status](https://status.twilio.com/)). - **Cloudflare One**: Some Cloudflare One Client users are being wrongly shown challenges on some sites. A fix is being rolled out. **Everything else is normal** AWS, Google Cloud, Azure, Azure DevOps, GitHub, OpenAI, Anthropic (Claude), Discord, Atlassian, Slack, Vercel, Heroku, DigitalOcean, Stripe, Cursor, Groq, ElevenLabs, Cohere and Stability AI all report no active incidents. Crowd-sourced trackers show no broad spikes, and IODA's country-level outage signals over the last 24 hours are limited to small local events. **Recently resolved** Facebook and Instagram had a short disruption for US users on Sunday night (September 20, US time). Meta did not comment on it and service returned quickly ([Tom's Guide](https://www.tomsguide.com/news/live/facebook-instagram-down-meta-outage-9-20-2026)). WhatsApp was reportedly unaffected. **Bottom line:** The internet is up. Expect slower routes across Asia-Pacific while cable capacity is restored.

No internet-wide outage is in progress as of about 12:00 UTC on September 23. Core cloud platforms (AWS, Google Cloud, Azure, Azure DevOps) show no active incidents. Several smaller provider incidents are open, and most are recovering. **GitHub: API and Projects disruption (recovering)** Starting around 10:11 UTC, GitHub reported degraded API performance. It traced the problem to detached database replicas, which also affected organization creation and GitHub Projects. Replicas were restored and the API degradation was mitigated by 10:58 UTC. At 11:38 UTC Projects was recovering, but search results may be stale while indexing catches up. The incident is still open. Source: https://www.githubstatus.com **Cloudflare: residual R2 and API authentication errors** Cloudflare says a small percentage of requests to its API and R2 object storage failed authentication. The issue started on September 22 at 13:30 UTC, and the main impact was mitigated by 19:00 UTC. Cloudflare is still working on the remaining errors. Separately, Cloudflare One Client (WARP) users are being wrongly challenged on some websites; a fix is being rolled out. Also in the past 24 hours, all resolved: elevated Durable Objects errors (08:09 to 10:00 UTC today), R2 503 errors in Sydney (about 17 minutes, 04:08 UTC), an issue with 1.1.1.1 for Families (00:42 to 01:26 UTC), a five-minute performance dip in Taipei, and brief Bot Management and Stream Live problems on September 22. Source: https://www.cloudflarestatus.com **Twilio: SMS delivery failures from alphanumeric sender IDs** Since about 10:58 UTC, Twilio has reported SMS delivery failures from alphanumeric sender IDs to several networks in several countries. It has identified the cause and is working on a fix. Some smaller carrier-specific delays are also open, including Tele2 Sweden and Jazz Pakistan. Source: https://status.twilio.com **Akamai: APAC latency linked to subsea cable work** Akamai reported elevated errors on September 22 when requests from outside Asia-Pacific reached origin servers in the region. It linked this to upstream congestion caused by unexpected maintenance on subsea cables and is monitoring the fix. Separately, planned upstream maintenance on routes between Singapore, Hong Kong and Japan runs until October 3, and some customers may see higher latency. This follows a series of subsea cable faults in the region this month, including damage to the Australia Singapore Cable and several of Vietnam's international cables. Source: https://www.akamaistatus.com **Everything else** Slack, Zoom, Atlassian, Vercel, Heroku, DigitalOcean, Stripe, Discord, OpenAI, Anthropic (Claude), Cursor, Groq, ElevenLabs, Cohere and Stability AI report normal operation. Zoom has planned Zoom Phone maintenance in its EU AMS2 data center. Fastly's status page blocks automated checks, and we saw no outage reports for it. We could not get machine-readable status for Meta (Facebook, Instagram, WhatsApp) or Replicate, and found no outage reports for either in search results. **Bottom line:** The internet is up. Today's incidents are partial and limited to specific services, and most have already been mitigated.

No widespread internet outage today. As of about 09:15 UTC on September 22, core cloud, CDN and DNS infrastructure is operating normally. Two notable but short incidents overlapped on Sunday evening US time, and both have recovered. Facebook and Instagram stopped loading for many US users from around 8:55 PM ET on Sunday, September 21 (00:55 UTC Monday). Downdetector logged more than 17,000 Facebook reports and over 5,000 for Instagram at the peak, and some users outside the US were affected too. Reports fell sharply by about 10 PM ET and service was normal by Monday morning. Meta has not given a cause, and its business status page at https://metastatus.com/ shows nothing open. This is the third notable Meta disruption of 2026, after larger ones in June and July. AWS had a control-plane problem in its busiest region at the same time. Between 22:49 and 00:06 UTC (3:49 to 5:06 PM PDT on September 21), new EC2 instance launches in us-east-1 (N. Virginia) failed or returned errors through RunInstances, CreateFleet and related APIs. Services that launch instances on customers' behalf, including ECS, Fargate and EMR, were also affected. Already-running instances were not impacted. AWS attributes it to a recent deployment to a subsystem that processes instance launches, and marks the event resolved. Details at https://health.aws.amazon.com/health/status Cloudflare closed two brief, minor edge incidents this morning: network connectivity in London (LHR) from 02:25 to 03:36 UTC and performance in San Jose (SJC) from 07:20 to 08:20 UTC. On Sunday afternoon, R2 PUT requests saw elevated failures in the WNAM and ENAM regions from 15:25 to 18:00 UTC, and Containers briefly could not start in Asia-Pacific. The low-impact WARP geolocation issue open since August 27 is still under repair. See https://www.cloudflarestatus.com/ One item is still open. Azure DevOps reports Test Plans degraded in Europe, with Microsoft "investigating a service disruption" as of 09:00 UTC. Core services, Repos, Pipelines and Boards are healthy in every region, and the main Azure status page shows no active events. Track it at https://status.dev.azure.com/ The AI platforms had a busy night but are clean now. Anthropic reported elevated error rates across several Claude models from 00:50 to 02:10 UTC, resolved after the root cause was found (https://status.claude.com/). Cursor resolved a degradation affecting Grok model requests at 01:42 UTC, its second Grok-related incident in two days. OpenAI, Groq, Replicate, ElevenLabs, Cohere and Stability AI report no incidents. Smaller items: Akamai has an advisory for elevated latency in Singapore, Hong Kong and Japan during upstream provider maintenance that runs through October 3; core delivery is unaffected. Twilio lists SMS delivery delays to Airtel in India, Du in the UAE and Telemat in Brazil, plus continuing voice call failures to Hong Kong, all carrier-level rather than platform-wide. Zoom resolved an Apple sign-in problem for a subset of users on Sunday. Earlier in the 72-hour window, GitHub had delays creating merge commits for about an hour on Saturday (22:13 to 23:22 UTC), and Vercel cleared three deployment-related incidents on Thursday. All clear at Google Cloud, Azure, GitHub, Discord, Atlassian, Slack, Vercel, Heroku, DigitalOcean and Stripe. No BGP or DNS-level events of note, and no new submarine cable damage reported. Bottom line: two short, contained incidents on Sunday evening, one minor regional item open at Azure DevOps, nothing internet-wide. The internet is up.

No widespread internet outage today. As of about 09:30 UTC on September 18, the major cloud platforms, CDNs and AI providers are all reporting normal service, and the incidents of the past 24 hours were short and are already closed. Cloudflare accounted for most of the activity. Three separate incidents opened and closed in roughly 24 hours: a network performance problem in Ashburn, Virginia (18:29 to 19:58 UTC on September 17), elevated errors for Durable Objects and downstream services in APAC including D1, Containers, Workers Assets and Workflows (23:37 UTC September 17 to 00:54 UTC September 18), and a Workers KV problem affecting the Cloudflare dashboard this morning (06:42 to 08:15 UTC). All three are marked resolved. One older, low-impact item is still open: some WARP users have been getting an incorrect geographic location since August 27, and Cloudflare says a fix is still in progress. Details on the Cloudflare status page: https://www.cloudflarestatus.com/ Reddit had a rough few hours overnight. Crowd-sourced trackers logged a spike in reports around midnight US Eastern, mostly mobile app failures, login errors and feeds that would not load, plus a short feed problem around 06:42 UTC. Reddit never posted an incident of its own, and https://www.redditstatus.com/ shows all systems operational, so treat the scope as unconfirmed. It is working normally now. Everything else was routine. GitHub cleared elevated errors for OpenAI models in Copilot on September 17, and Cursor resolved a degradation in Cloud Agents and Automations the same day. Discord fixed call connection delays, and Vercel closed a log drain delivery failure. Twilio is the one carrier-level item still showing degradation, with slower Voice and SMS in APAC, a Bulk Export issue, and continuing voice call failures from a subset of numbers to Hong Kong. Zoom Phone in North America is under scheduled maintenance, which is planned work rather than a fault. The core platforms are clean. AWS, Google Cloud, Microsoft Azure, Azure DevOps, Akamai, OpenAI, Anthropic, Atlassian, Slack, Heroku, DigitalOcean, Groq, ElevenLabs, Cohere and Stability AI all report no active incidents. Worth noting from the past 72 hours: Microsoft confirmed a Microsoft 365 service degradation on September 16 (tracked as MO1472904), where users hit 502 and 503 errors across Outlook, Teams and other apps. That has since been restored and Microsoft's status page shows no active events. Akamai and DigitalOcean both had support ticketing outages on September 16 caused by an upstream provider, which affected help desks rather than production traffic. At the network layer there is nothing unusual. Country-level monitoring from Georgia Tech's IODA project shows only small, localized dips in places such as Tajikistan, Liberia, Bermuda and Turks and Caicos, all consistent with ordinary regional variation rather than a national shutdown. No new BGP hijacks or route leaks of note, and no fresh submarine cable damage has been reported. One correction worth making, because it is circulating: stories about ChatGPT, Claude and Gemini going down simultaneously describe an incident on September 8, not today. Both https://status.openai.com/ and https://status.claude.com/ currently report full availability. Bottom line: a handful of brief, contained incidents, no sign of anything internet-wide. The internet is up.

No widespread internet problems were detected as of 10:30 UTC on September 17, 2026. Core infrastructure (Google Cloud, Azure, Cloudflare, GitHub, Akamai, OpenAI, Anthropic, Discord, Atlassian, Slack, Zoom, Vercel, Heroku, DigitalOcean, Stripe) is reporting normal operations, and crowd-sourced outage trackers show no unusual spikes this morning. **AWS confirms permanent data loss in its Bahrain region** The most significant development of the past 72 hours is a pair of updates AWS posted on September 15 to its long-running Middle East disruption. For the Bahrain region (me-south-1), AWS says the damage from the March conflict "spanned multiple Availability Zones and exceeded what our regional and multi-AZ services are designed to withstand," and that it is "unable to restore access to the resources and data hosted exclusively in this Region." The next update is promised for early 2027. For the UAE region (me-central-1), AWS says resources and data hosted only in the mec1-az2 zone cannot be recovered, while recovery work continues on the other two zones. Both regions have been unavailable or unreliable since March 2026, and AWS has been steering customers to other regions since then. Source: https://health.aws.amazon.com/health/status **Resolved on September 16** - **Discord** had a roughly 50-minute voice-call outage in the afternoon US time, with thousands of Downdetector reports peaking around 3 PM ET, following a shorter API-errors incident earlier in the day. Both are resolved. Source: https://discordstatus.com - **Google Play billing** had problems with purchases and subscription management for several hours. The knock-on effect showed up on other companies' status pages: Discord could not complete Android purchases, and Anthropic reported failed Claude subscriptions on Android. Both cleared once Google fixed the underlying issue. Sources: https://discordstatus.com · https://status.claude.com - **Cloudflare** resolved three small network-level issues: HTTP 522 errors from its St. Louis facility, degraded performance in Los Angeles, and a 1.1.1.1 resolver problem with Tunisian (.tn) domains. Source: https://www.cloudflarestatus.com - **GitHub Copilot** saw degraded performance from third-party model providers on September 15 and 16, and Anthropic reported intermittent error spikes for its Mythos 5.1 and Fable 5.1 models on September 15. All resolved. Sources: https://www.githubstatus.com · https://status.claude.com - **Akamai** had a 12-minute HTTPS edge-delivery performance issue early on September 15. Source: https://www.akamaistatus.com **Minor items still open** - **Twilio** lists a batch of carrier-specific SMS, MMS and voice delivery issues opened overnight (Optus and other Australian carriers, Three UK, Hong Kong, GCI in the US, Rogers in Canada) plus an SNA authentication issue for PLDT Smart in the Philippines. These are narrow, per-carrier problems of the kind Twilio reports routinely. Source: https://status.twilio.com - **Replicate** reports some Pruna models unable to scale up, causing long queues or unavailability. Source: https://www.cloudflarestatus.com - **Cloudflare WARP** continues to report incorrect geolocation for some users, an issue open since August 27. Crowd reports also mention IndiHome and Telkomsel in Indonesia around midnight local time on September 17. We could not confirm these with an official source, and volumes appear modest. **Bottom line:** the internet is up. Yesterday's Discord and Google Play disruptions are over, and the AWS announcement is a formal write-off of infrastructure that has been unavailable for months rather than a new outage.

No widespread internet disruption is in progress. As of roughly 09:30 UTC on September 14, the major cloud platforms, CDNs and DNS infrastructure are operating normally. Google Cloud reports no broad severe incidents (https://status.cloud.google.com), Azure shows no active events (https://azure.status.microsoft/en-us/status), and Akamai (https://www.akamaistatus.com) and GitHub (https://www.githubstatus.com) are fully operational. Cloudflare is carrying traffic normally apart from routine datacenter maintenance in Dallas, Atlanta, Ashburn and Newark (https://www.cloudflarestatus.com). Active incidents: - Twilio is investigating SMS delivery delays to Vivo subscribers in Brazil and voice call failures to Hong Kong numbers. Both are carrier-level problems rather than platform-wide ones (https://status.twilio.com) - Slack has a long-running bug affecting access to historical messages in workspaces that use custom data retention policies. Open since August 13; other Slack functions are normal (https://slack-status.com) - Cloudflare still lists a minor issue reporting incorrect geolocation for some WARP users, open since August 27 (https://www.cloudflarestatus.com) - Anthropic's Claude Cowork remains degraded on Windows after a September 8 Windows update removed its access to workspace drives. Chat and file features still work, and Microsoft has a fix pending (https://status.claude.com) AWS Middle East regions remain offline: The most significant ongoing infrastructure problem is not new. AWS ME-CENTRAL-1 (UAE) and ME-SOUTH-1 (Bahrain) have been unable to reliably support customer workloads since drone strikes damaged data centers there in March, and AWS has said full restoration will take several months. Customers with resources in those regions are still advised to migrate elsewhere or restore from remote backups, and billing operations there are suspended. Every other AWS region is operating normally (https://health.aws.amazon.com/health/status). Regional connectivity: IODA has recorded critical-level connectivity loss around Kherson, Ukraine, consistent with the power outages reported across the Kherson and Mykolaiv areas this month. This is a local electricity and infrastructure problem rather than an internet routing failure, and it does not affect traffic outside the region (https://ioda.inetintel.cc.gatech.edu/country/UA). Resolved in the past 72 hours: - GitHub had roughly 90 minutes of degraded Pull Requests and Actions on September 13, caused by database replication delays that pushed up error rates on authorization endpoints. Netlify and CircleCI reported degradation in the same window (https://www.githubstatus.com) - OpenAI had two short periods of degraded service on September 13, lasting 27 and 51 minutes (https://status.openai.com) - Stripe cleared elevated Interac errors on September 13 after partner maintenance completed (https://status.stripe.com) - Akamai resolved an Edge Delivery latency issue in US East on September 11, and Discord resolved high media proxy latency the same day (https://discordstatus.com) - X saw a spike in user outage reports on the evening of September 13 that subsided without an official statement. X does not publish a machine-readable status page, so this rests on user reports alone One item we could not confirm: a modest number of Microsoft 365 user reports appeared in Japan early on September 14. Microsoft's own status pages show no active events, so this appears to have been minor or short-lived rather than a broad outage. Overall: a quiet day. Nothing currently in progress would affect most people's ability to use the internet.

Most of the internet ran normally today. The one significant event was at GitHub, which had a critical incident between 09:16 and 10:44 UTC. Pull Requests were fully unavailable for part of that window, and API Requests, Issues, Pages and Actions were degraded. GitHub attributed it to increased database replication delays on its collab database, which pushed error rates up on authorization endpoints and then spread across dependent services. Engineers shed load and brought services back gradually. The incident is resolved. The large cloud and edge providers were quiet. AWS, Google Cloud, Azure, Akamai, DigitalOcean, Heroku and Vercel all reported normal operation. Cloudflare has one open minor issue, incorrect geo-location for WARP users, dating to August 27, and has datacenter maintenance windows scheduled September 14 through 18 in Ashburn, Newark, Phoenix, Atlanta, Chicago, Portland, Sao Paulo, Chennai and Rome. Traffic may be rerouted during those windows with a small latency increase. AI platforms were stable. OpenAI, Groq, Cohere, ElevenLabs, Stability AI and Cursor all reported no issues. Anthropic has one open item unrelated to service availability: a Windows update released September 8 prevents Claude Cowork on Windows from running local commands. Chat and file operations still work. Twilio carries three open carrier-level delivery problems: SMS failures from some Twilio long codes to Bouygues Telecom subscribers in France, MMS delivery trouble on some Australian networks, and voice call failures to some Hong Kong numbers. These affect specific routes rather than the platform as a whole. Slack continues to list a long-running issue retrieving historical messages for workspaces with custom data retention policies. One regional signal is worth flagging with a caveat. IODA detected a connectivity drop in Ghana starting around 10:55 UTC, concentrated in the Brong Ahafo region, and it was still showing as active at 12:00 UTC. The detection sits at 65 percent confidence from a single measurement source with no corroborating reports, so treat it as unconfirmed. No cause has been established and nothing suggests it extends beyond Ghana. Looking back over the past 72 hours, everything notable has closed out: Cloudflare resolved Tunnel availability problems and Workers Cron Trigger degradation on September 12, Akamai cleared increased Edge Delivery latency in US East on September 11, Discord fixed media proxy errors on September 11, and Vercel recovered from elevated errors on Connect and Passport on September 10. No BGP hijacks, route leaks or submarine cable faults were reported today, and Meta platforms showed no disruption. Overall: incidents detected, none of them internet-wide. Sources: GitHub Status (githubstatus.com), Cloudflare Status (cloudflarestatus.com), Twilio Status (status.twilio.com), Claude Status (status.claude.com), IODA (ioda.inetintel.cc.gatech.edu).

No internet-wide outage today. Core routing, DNS and the major cloud platforms are carrying traffic normally. The notable item is an unresolved Slack incident affecting stored files, plus a handful of smaller provider faults that have mostly already cleared. Slack (ongoing). Since 11:04 UTC, a subset of Slack customers cannot load or download images uploaded before roughly 21:30 UTC on September 11. Affected files show a blurry preview and fail to download across desktop, web and mobile. Newly uploaded images work normally. Slack's first theory was a thumbnail-sharpening change and rolled it back, which did not help; the team has since confirmed the trigger as a code change that caused older file records to lose their stored file names, breaking retrieval. A targeted fix is being prepared and Slack expects affected files to recover automatically once it ships. No resolution estimate has been given. Slack also still has a separate long-running issue, open since August 13, affecting historical messages in channels with custom data retention policies. https://slack-status.com Cloudflare. A busy night, now largely clear. Cloudflare Tunnel availability problems ran from 21:55 UTC on September 11 to 08:57 UTC today, about eleven hours, and were the most substantial disruption in the window. Separately, Cloud Connector rule creation and updates failed between 08:06 and 10:42 UTC, and degraded Workers Cron Triggers resolved at 03:34 UTC after starting on September 9. All three are resolved. One minor incident stays open: incorrect geolocation for some WARP users, identified on August 27 and still being worked. https://www.cloudflarestatus.com Twilio. Two regional faults are open. Voice calls from a subset of Twilio numbers to Hong Kong subscribers are failing, with a cause identified and a fix in progress. MMS delivery from Australian networks to Twilio AU numbers is under investigation. Both are carrier-scoped rather than platform-wide. https://status.twilio.com Anthropic. Claude Cowork on Windows is in partial outage. A Windows update released September 8 stops it from running local commands or reaching the local drive; chat and file operations still work. Microsoft is developing the fix. Other Claude services are operational. https://status.claude.com AWS Middle East, ongoing background. Worth stating plainly because it makes third-party dashboards show AWS in a degraded state: the ME-CENTRAL-1 (UAE) and ME-SOUTH-1 (Bahrain) regions have been unavailable since early March after sustaining damage from the conflict in the Middle East. AWS advises customers to recover resources in other regions from backups and says restoration will take several months. This is not new today and does not affect AWS regions elsewhere, which are operating normally. Everything else checked clean. Google Cloud reports no broad severe incidents. Azure shows no active events. GitHub, OpenAI, Akamai, Atlassian, Zoom, Vercel, DigitalOcean, Discord and Meta's platforms are all operational, as are Cursor, Groq, Replicate, ElevenLabs, Cohere and Stability AI. Resolved in the past 72 hours. Akamai cleared increased edge latency in US East (IAD) early on September 11. Cloudflare resolved network performance problems at its Taipei datacenter and delays provisioning Let's Encrypt certificates, both on September 11. Discord fixed image download errors on its media proxy and voice call connection trouble on September 10 and 11. Vercel resolved elevated errors on Connect and Passport on September 10. DigitalOcean restored power redundancy at its ATL1 datacenter on September 9 after reduced capacity affected some GPU workloads, with no server outages. One unconfirmed signal. A modest number of Virgin Media users in the UK reported connectivity and sign-in problems overnight into today. Report volumes stayed low and Virgin Media has not acknowledged an outage, so we are noting this rather than calling it one.

Status as of 10:15 UTC on September 11, 2026: incidents detected, no widespread internet disruption. **Ongoing** - **OpenAI — ChatGPT errors in Europe.** OpenAI opened an incident at 08:37 UTC for elevated errors affecting ChatGPT users in Europe. Downdetector-driven coverage described users unable to load ChatGPT, generate images, or see conversation history. A mitigation was applied and the incident is in the monitoring phase; service is largely back with some sporadic slowness reported. A second incident, elevated errors for the GPT-5.6 Sol model on the API, opened at 09:08 UTC and is still marked identified with a mitigation in progress. Source: https://status.openai.com/ - **Akamai — increased edge latency in US East (IAD).** Opened 17:37 UTC on September 10, fix implemented 18:10 UTC, still in monitoring. All other Akamai services operational. Source: https://www.akamaistatus.com/ - **Cloudflare — minor degradations.** Workers Cron Triggers remain degraded (open since September 9, fix being implemented) and the WARP geolocation bug from August 27 is still open. Source: https://www.cloudflarestatus.com/ - **Twilio — carrier-level delivery issues** in Australia (MMS), Italy (SMS via Iliad) and Indonesia (SNA authentication), all in monitoring with recovery observed. Source: https://status.twilio.com/ - **Slack** continues to work on the custom data retention bug that blocks loading of some historical messages (open since August 13). Source: https://slack-status.com/ - **Anthropic** has an open issue with Claude Cowork on Windows, where a September 8 Windows update broke local command execution. Chat and file operations are unaffected. Source: https://status.claude.com/ **Resolved in the past 72 hours** - **Cloudflare DNS update delays**, September 10, 13:58 to 22:22 UTC (about 8.4 hours): DNS record changes propagated slowly. A shorter delay in Let's Encrypt certificate provisioning followed overnight, 02:20 to 04:05 UTC on September 11. - **Fidium Fiber** (Consolidated Communications' fiber ISP) had a regional broadband outage across New Hampshire, Maine and Vermont on September 10, roughly 2 hours 50 minutes, with more than 2,500 Downdetector reports at the peak. Source: https://gvwire.com/2026/09/10/fidium-fiber-goes-down-for-thousands-downdetector-reports/ - **Anthropic** resolved elevated latency and timeouts on the Claude API for US Midwest traffic at 23:24 UTC on September 10. - **Discord** fixed a voice-calling problem overnight (04:21 to 06:42 UTC on September 11). - **Zoom** had a brief four-minute degradation of Phone, Contact Center and Meetings in Europe on September 10 (14:58 to 15:02 UTC), plus a longer US-region issue with meeting and call summaries (17:57 to 23:49 UTC). - **Vercel** resolved elevated errors on Connect and Passport on September 10 (19:16 UTC). - **DigitalOcean** resolved a power degradation affecting GPU workloads in its ATL1 data center on September 9 (10:47 to 12:25 UTC). - **Cursor** resolved a 25-minute degradation of Review Agents and Cloud Agents on September 10. **Core infrastructure** AWS, Google Cloud, Microsoft Azure, Azure DevOps, GitHub, Heroku, Stripe, Atlassian, Groq, Replicate, ElevenLabs, Cohere and Stability AI all report normal operations with no incidents in the past 72 hours. AWS's two long-running Middle East region events (ME-CENTRAL-1 and ME-SOUTH-1, damaged in the regional conflict since March) remain open but are unchanged. No BGP route leaks, DNS root issues or undersea cable faults were reported today. IODA flagged only short, single-source anomalies overnight (a 10-minute blip for Washington state and a 6-minute blip for Iowa around 03:11 to 03:42 UTC) with no corroboration from ISP report volumes; Frontier and AT&T report counts are at baseline. **Unconfirmed** A news report published at 09:49 UTC said X (Twitter) was down across platforms. Independent report trackers show X at baseline volumes and the site is loading normally as of this writing, so this appears to have been brief or limited if it occurred at all. **Bottom line:** the internet is up. The morning's notable disruption is OpenAI's European ChatGPT errors, now mitigated, alongside a handful of minor provider-level issues. Related: yesterday's report covered the Microsoft Copilot outage: https://www.isinternetup.com/updates/2026-09-10

No widespread internet disruption today. The notable item is a Microsoft Copilot outage that ran overnight and has since been fixed. Microsoft Copilot (resolved). The Copilot website was unreachable from 22:25 UTC on September 9 until 00:05 UTC on September 10, about an hour and 40 minutes. Users got a Cloudflare Error 1016 instead of the chat interface. Downdetector logged a peak of roughly 4,200 reports. Microsoft traced it to an affected network flow and applied a configuration fix. Separately, and at the same time, an internal Microsoft 365 resilience drill removed the suggested-prompt pills in Copilot Chat. Microsoft halted the drill once it spotted the impact. Both issues are resolved. Still open, all minor: - Cloudflare has two low-impact incidents on its board: Workers Cron Triggers degraded (opened September 9, fix being implemented) and a WARP geolocation issue open since August 27. Scheduled datacenter maintenance is running in Amsterdam, Los Angeles, Phoenix and Saint Petersburg, which may add latency in those metros. - Twilio reports SMS delivery receipt delays from a subset of long codes to Iliad subscribers in Italy, opened this morning. Messages are being delivered; it is the receipts that lag. An MMS delay issue in Australia has been open since September 4. - OpenAI is monitoring recovery on a bug that blocked opening shared ChatGPT Projects from a direct link. Mitigation is applied. Core ChatGPT and API service is normal. - Slack still has an incident open since August 13 affecting workspaces with custom data retention policies, where older messages may fail to load. Operating normally: AWS, Google Cloud, Microsoft Azure, Cloudflare core network, GitHub, Akamai, Anthropic, Discord, Atlassian, Zoom, Vercel, Heroku, DigitalOcean, Stripe, Cursor, Groq, Cohere and Stability AI. We found no BGP hijacks, route leaks or DNS failures of note. Recently resolved (past 72 hours). AT&T had an internet disruption on September 7 that peaked near 5,900 reports and affected Dallas, Phoenix, New York, Philadelphia, Chicago and Los Angeles. AT&T attributed the Dallas portion to a fiber equipment issue and confirmed service returned to normal the same day. On September 4, GitHub resolved a Copilot code review failure and a repository contents API degradation, and Akamai resolved a load balancer issue plus an Atlanta network component failure that affected certificate provisioning. A note on sources. Several search results circulating today describe Red Sea submarine cable cuts near Jeddah affecting the SMW4 and IMEWE systems. Those reports date from September 2025, not this week, and we found no evidence of a new break. AWS public RSS feed also continues to carry long-running Middle East regional entries that have not been updated in months, so treat that feed timestamps with care. Overall: incidents detected, none widespread. The internet is up. Sources: Microsoft Copilot outage coverage via The Register (https://www.theregister.com/ai-and-ml/2026/09/10/ai-uprising-postponed-after-copilot-falls-off-the-web/5295475) and GV Wire (https://gvwire.com/2026/09/09/microsoft-copilot-goes-down-for-thousands-downdetector-reports/); AT&T via Newsweek (https://www.newsweek.com/is-att-down-thousands-of-users-report-problems-with-service-12413362); provider status pages at https://www.cloudflarestatus.com/, https://status.twilio.com/, https://status.openai.com/, https://slack-status.com/, https://www.githubstatus.com/, https://www.akamaistatus.com/, https://status.cloud.google.com/ and https://azure.status.microsoft/.

No internet-wide outage today. The backbone and the major clouds are healthy, but several named provider incidents were open as of 19:30 UTC. Cloudflare has two open incidents, both rated minor impact. Workers AI has shown increased error rates since 14:25 UTC, and Workers Cron Triggers have been degraded since 19:17 UTC. Both sit at "identified" with a fix being implemented, so scheduled Workers jobs and AI inference calls may fail or run late until they close. A separate burst of HTTP errors at Cloudflare's Philadelphia location ran 16:40-16:54 UTC and is fully mitigated. Details: https://www.cloudflarestatus.com/ Twilio has several carrier-level problems open at once: SMS delivery failures from certain long codes to AT&T subscribers in the US (identified), voice call failures to Verizon in the US (recovering, under monitoring), and MMS delivery delays to Australian numbers (under investigation). These are last-mile carrier issues rather than a Twilio platform outage, but they affect real message and call delivery. Details: https://status.twilio.com/ Consumer outage trackers logged report spikes at several North American providers through the day, none of which the providers have confirmed as a broad network failure: Rogers in Canada (surges near 02:13 and 11:13 ET), CenturyLink (07:37 ET), Frontier (11:45 ET), plus YouTube (03:35 ET) and Amazon's retail site (10:52 ET). Spikes of this size are usually regional or short-lived. One unconfirmed signal is worth naming rather than burying. IODA flagged a critical-level connectivity anomaly for Iowa at 13:20 UTC, measuring roughly an 85% drop against baseline. That reading comes from a single measurement source with low corroboration, so treat it as unverified. Drake University in Des Moines separately reported brief, repeated internet interruptions the same day, which is consistent with a localized problem but does not establish its scale. Smaller items: DigitalOcean lost power redundancy in its ATL1 datacenter, briefly affecting GPU droplet performance, and has since restored it. ElevenLabs is investigating a broken MCP connection in ElevenCreative, with ElevenAgents unaffected. Slack is still working an issue open since August 13 in which workspaces with custom data retention policies have trouble loading historical messages. Zoom's September 8 degradation of Phone, Contact Center and Meetings is resolved and being monitored. AWS's Middle East regions remain the longest-running item in cloud infrastructure. ME-CENTRAL-1 (UAE) and ME-SOUTH-1 (Bahrain) have been unavailable since physical damage earlier this year, and AWS still advises recovering those workloads from backups held in other regions. AWS's public feed updates slowly, so the absence of a new entry there is not by itself proof of an all-clear. Resolved in the past 72 hours: AT&T had a 57-minute backbone disruption centered on nodes in Ashburn, Virginia early on September 6. Akamai fixed an Application Load Balancer availability problem and a certificate provisioning issue tied to Atlanta network connectivity on September 4. GitHub resolved elevated failures in Copilot code review and a Repository Contents API degradation, also on September 4. Cursor cleared login and Grok Bot errors on September 7. Everything else we checked reports normal operation: Google Cloud, Azure, GitHub, Akamai, OpenAI, Anthropic, Discord, Atlassian, Vercel, Heroku, DigitalOcean, Groq, Replicate, Cohere and Stability AI. Meta's platforms are serving normally, with a modest trickle of user reports about slow loading rather than hard failures. Two checks were inconclusive: Stripe's and Meta's own status pages did not render for our automated request, so neither is independently confirmed here. Bottom line: the internet is up. If you are seeing failures today, the most likely candidates are Cloudflare Workers, SMS and voice delivery through Twilio to AT&T or Verizon, or a regional ISP problem at your own provider.

No widespread internet disruption is in progress. The major cloud platforms, CDNs and developer services are all reporting normal operation this morning, and the one significant consumer outage of the past 72 hours - AT&T in Texas - has been resolved. Open incidents Slack has an ongoing degradation affecting access to historical messages for workspaces that have custom data retention policies enabled. The incident has been open since August 13 and still shows Messaging and Workspace/Org Administration as impaired. Workspaces on default retention settings are not affected. (https://slack-status.com/) Twilio has three open carrier-specific delivery issues: SMS delivery delays to Megafon Tajikistan, delayed delivery receipts from Twilio long codes to Nex-Tech Wireless in the US, and MMS delivery delays from Australian subscribers to Twilio Australia numbers. Twilio says it has identified the cause in each case. Message delivery itself is largely succeeding; the delays are in receipts and specific carrier paths. (https://status.twilio.com/) Cloudflare lists one minor open issue - incorrect geolocation for some WARP users - alongside routine scheduled datacenter maintenance in Philadelphia, Dallas, Berlin and Phoenix. Traffic is re-routed during those windows, so a small latency increase in those metros is expected and is not a fault. (https://www.cloudflarestatus.com/) AWS continues to list the Middle East (UAE and Bahrain) regional disruption that began in April 2026. AWS still advises customers to migrate workloads out of those regions. This is long-running background rather than a new event, and no other AWS region is affected. All clear elsewhere Google Cloud, Azure, Azure DevOps, GitHub, Akamai, OpenAI, Anthropic, Discord, Atlassian, Zoom, Vercel, Heroku, DigitalOcean, Stripe, Cursor, Groq, Replicate, ElevenLabs, Cohere and Stability AI all report normal operation. Recently resolved AT&T customers across Texas, concentrated in Houston and Dallas-Fort Worth, reported widespread broadband, fiber and 5G home internet problems on Monday, September 7. Reports peaked at roughly 17 times normal volume around midday. AT&T said service had been restored by late Monday afternoon. (https://www.wfaa.com/article/news/local/att-outage-dallas-texas/287-d12d4412-abfa-444e-b3ea-0caf32407ca9) Akamai resolved two incidents on September 4 - an Application Load Balancer availability issue and a network component failure in Atlanta that affected certificate provisioning. GitHub resolved Copilot Code Review and repository contents API degradations on September 4. Zoom resolved an AI Companion degradation in the US region on September 5. One unconfirmed signal A third-party aggregator flagged a critical connectivity anomaly for Washington state at 00:05 UTC today, attributed to Georgia Tech's IODA project. IODA's own alert feed returns no US alerts for the past 24 hours, and the aggregator's own page rates the finding as low-corroboration from a single data source. We are noting it rather than reporting it as an outage, because we could not verify it against the source it cites. Overall: the internet is working normally. A handful of narrow, provider-specific issues are open, none of them broad enough to affect general connectivity.

No internet-wide outage is in progress. The major cloud and CDN platforms are reporting normal service, and the issues we can confirm today are narrow ones affecting specific features rather than whole networks. **Active, provider-confirmed** Twilio is the most substantial open incident. The company reports MMS delivery delays and failures from a subset of its US short codes to multiple carrier networks, with the cause identified and a fix in progress. Separately, Twilio lists MMS delivery failures on Telstra in Australia affecting a subset of Australian Twilio numbers, and is monitoring recovery on SMS delivery to Robi subscribers in Bangladesh. This is a messaging-delivery problem, so the visible effect is one-time passcodes, alerts and picture messages arriving late or not at all on affected routes rather than any loss of connectivity. Source: https://status.twilio.com/ Cloudflare is carrying three minor incidents, none of them affecting content delivery or DNS resolution. A Networking Routes UI problem in the dashboard began at 15:32 UTC and a fix is being implemented; Browser Isolation session failures affecting Browser Isolation and Gateway started at 14:25 UTC and are in monitoring after a fix; and a WARP geolocation error open since 27 August is still being worked on. Source: https://www.cloudflarestatus.com/ Slack continues to carry a long-running minor incident, open since 13 August, in which customers with custom data retention policies enabled have trouble accessing historical messages. Everything else on Slack's board is clear. Source: https://slack-status.com/ **Consumer ISP reports** Crowd-sourced trackers logged report spikes at two large consumer ISPs in the early hours, both of which have since settled. AT&T reports climbed from roughly 03:20 ET, weighted toward 5G Home Internet, broadband and Wi-Fi, with complaints concentrated in Texas including a fibre fault in north Dallas. Virgin Media in the UK saw reports rise from roughly 07:00 BST, weighted toward landline internet and email. Neither company published a status-page notice, and by the afternoon istheservicedown showed no problem detected at AT&T, so these look like regional faults rather than national outages. Report-volume data is a crowd signal, not a measurement, and we are treating the scope of both as unconfirmed. **Clear on checks** AWS, Microsoft Azure, Google Cloud, GitHub, Akamai, Discord, Atlassian, Zoom, Vercel, Stripe, DigitalOcean and Heroku all report normal service. On the AI side, OpenAI, Anthropic, Cursor, Groq, Replicate, ElevenLabs, Cohere and Stability AI are all operational. Meta's status page renders client-side and could not be read directly, but no Facebook, Instagram or WhatsApp disruption surfaced in outage reporting today. No large-scale BGP hijack, route leak or submarine cable incident was reported. **Recently resolved (past 72 hours)** Cursor had login and account errors between 12:12 and 13:08 UTC today, affecting the CLI, Cloud Agents and the IDE, now fixed. Cloudflare resolved a TURN service degradation and Workflows instance-creation errors earlier today, plus a Registrar display problem and connectivity issues in its Canberra location on 6 September. On 4 September, Akamai worked through an Application Load Balancer reporting and availability issue (06:15 to 14:20 UTC) and a network component failure in Atlanta that took down its Certificate Provisioning System for about two hours, while GitHub cleared a Copilot Code Review disruption and a repository contents API degradation. For wider context, the week's two larger events both fall outside the 72-hour window: a Google Cloud zonal failure in us-central1-b on 1 September that touched 15 products for just over four hours, and a roughly 90-minute window on 3 September in which several major AI assistants degraded at once. **Assessment:** incidents detected, no widespread outage. Scope is single-provider and feature-level in each case, and the trajectory is stable to recovering.

No widespread internet disruption today. The major cloud platforms, CDNs and AI providers are all reporting normal operations, and there is no sign of routing, DNS or subsea cable trouble affecting traffic at scale. A few narrow incidents remain open: Slack - an issue affecting access to historical messages in workspaces with custom data retention policies has been open since August 13. Messaging and workspace administration are flagged, but Slack reports no broader disruption (slack-status.com). Cloudflare - some WARP users are being assigned an incorrect geographic location, open since August 27 and rated minor impact. Cloudflare also cleared several short-lived issues earlier this week, including elevated R2 errors in eastern North America on Sept 4-5 and a 16-minute connectivity degradation in Tokyo (cloudflarestatus.com). Twilio - three carrier-specific messaging problems: MMS delivery receipts to GCI in the US (now recovering), SMS delivery receipts to Mauritel in Mauritania, and MMS failures from Telstra in Australia. These affect individual carrier routes rather than Twilio as a whole (status.twilio.com). Everything else checked clean. AWS, Google Cloud, Azure, GitHub, Akamai, Atlassian, Vercel, DigitalOcean, Zoom, Discord and Cursor show no active incidents, as do OpenAI, Anthropic, Groq, Replicate, ElevenLabs, Cohere and Stability AI. Resolved in the past 72 hours: GitHub cleared a Copilot Code Review disruption and a repository contents API degradation on Sept 4. Akamai resolved an application load balancer fault and a certificate provisioning problem affecting Atlanta the same day. Zoom fixed roughly two hours of degradation affecting AI Companion in the US region on Sept 5. Discord resolved intermittent server availability problems on Sept 3. None have recurred. One clarification, because search results are currently muddled on it: the Red Sea subsea cable cuts near Jeddah that damaged SMW4, IMEWE, FALCON GCX and Europe India Gateway, and pushed Microsoft to reroute Azure traffic, happened on September 6, 2025 - one year ago today, not today. Azure routing was stabilised the following day. There is no new Red Sea cable incident at this time. The internet is up.

No widespread internet disruption is in progress as of 10:30 UTC on September 5, 2026. The overnight period was quiet, and the largest single event had already closed before the North American morning. Cloudflare R2 object storage returned elevated 503 errors in Eastern North America starting at 21:10 UTC on September 4. Cloudflare identified the cause within about 20 minutes and marked the incident resolved at 00:42 UTC on September 5, roughly three and a half hours end to end. A shorter R2 incident hit Western North America earlier on September 4, from 02:58 to 03:28 UTC. Both are closed. One minor Cloudflare issue stays open: incorrect geolocation for some WARP users, unresolved since August 27 and low impact. Details on the Cloudflare status page: https://www.cloudflarestatus.com/ Twilio is carrying three open incidents this morning, all narrow in scope: SMS delivery failures to Africell and Qcell in Gambia, an elevated authentication API failure rate affecting one identity gateway partner in the United States, and voice call failures from a subset of Twilio numbers to Switzerland. These affect specific carrier routes rather than the platform as a whole. See https://status.twilio.com/ Slack still lists a long-running issue with access to historical messages for workspaces on custom data retention policies, open since August 13 and unchanged. Everything else came back clean. AWS, Google Cloud and Azure all show no active events. GitHub, Akamai, OpenAI, Anthropic, Discord, Atlassian, Zoom, Vercel, Heroku, DigitalOcean, Stripe, Cursor, Groq, Replicate, ElevenLabs, Cohere and Stability AI are operational. Meta platforms are up. No BGP hijacks, route leaks or submarine cable faults of note were reported in the past 24 hours. Resolved in the past 72 hours: Akamai lost a network component in Atlanta on September 4, disrupting certificate provisioning and some edge traffic between 12:21 and 14:19 UTC, alongside a separate Application Load Balancer reporting fault the same day that ran about eight hours. September 3 brought a cluster of AI provider errors, with Anthropic, Cursor and GitHub Copilot each logging degraded model availability inside the same window; all recovered that evening. No common upstream cause has been confirmed by any of the three. Zoom cleared regional phone audio problems and North American SMS delays on September 3, and Discord resolved repeated server availability incidents on September 2 and 3. One standing note that predates this window: AWS me-central-1 (UAE) and me-south-1 (Bahrain) have been unavailable since April 30 following conflict damage, and AWS still expects recovery to take months. Nothing changed there today, but it remains the longest-running regional outage on the board. Verdict: incidents detected, none widespread. The internet is up.

The internet is broadly healthy this morning. One incident is still open at OpenAI, and yesterday's simultaneous outage across three major AI chatbots has been resolved. Open now: OpenAI is investigating elevated error rates for users in the Asia-Pacific region, affecting ChatGPT, Work mode, image generation, file uploads, Voice and Codex Cloud. The incident opened at 07:00 UTC and was still marked as investigating as of 09:30 UTC. Users outside APAC are not reported as affected. Source: https://status.openai.com/ Yesterday's AI outage, now resolved: on September 3, ChatGPT, Claude and Grok all failed at roughly the same time. Downdetector logged more than 37,000 reports for OpenAI, around 1,365 for Grok and about 1,300 for Claude. OpenAI described a routing error beginning around 07:43 PT that made ChatGPT and Codex unavailable for some users, and confirmed all services were back to normal by 12:38 PT. Anthropic's status page shows its model error incidents resolved by 16:23 UTC. Cursor was affected downstream through both its OpenAI and Anthropic model routes. Press coverage attributed the common cause to a regional failure in Microsoft Azure's East US infrastructure, which hosts capacity for all three; we note that Azure's public status page and status history list no corresponding event, so that attribution is not confirmed by the provider. Google's Gemini, which runs on Google Cloud, stayed largely available. Smaller items still open: Akamai is investigating a problem with Application Load Balancer reporting. Cloudflare has an unresolved issue returning incorrect geolocation for some WARP users, open since August 27, alongside a set of regional incidents from the past two days that are all now closed, including a hardware failure in Tokyo and R2 errors in western North America. Slack continues to show degraded messaging and admin components tied to retrieving historical messages in workspaces with custom data retention policies, an issue open since August 13. Twilio has several carrier-specific SMS and MMS delivery problems, including delays to Vivo and Claro subscribers in Brazil. Core infrastructure is clear. AWS reports no open issues, Google Cloud shows no broad severe incidents, and Azure lists no active events. GitHub, Discord, Atlassian, Vercel, DigitalOcean, Zoom, Cursor, Groq, Replicate, ElevenLabs, Cohere and Stability all report normal operation. Searches for BGP incidents, DNS failures and submarine cable faults surfaced nothing new; the only country-level disruptions are the long-running intentional shutdowns tracked by the Internet Society, which stood at 11 as of September 3. Overall: incidents detected, no widespread outage. Sources: https://status.openai.com/ · https://status.claude.com/ · https://www.cloudflarestatus.com/ · https://www.akamaistatus.com/ · https://slack-status.com/ · https://status.twilio.com/ · https://health.aws.amazon.com/health/status · https://status.cloud.google.com/ · https://azure.status.microsoft/en-us/status

No internet-wide outage today. Most major providers are running normally, with a few contained incidents and one large recent outage still finishing its recovery. Twilio is the main active incident. The company reported voice call failures from a subset of Twilio phone numbers to multiple countries in Europe, identified at 03:24 PDT, with engineers working on a fix. Separately, Twilio is tracking SMS and MMS delivery delays across several carriers: C-Spire and Dish in the US, Three in the UK, long code traffic in the Netherlands, and Algar Telecom in Brazil. Several of those show recovery underway. Message Logs also saw elevated latency on a small share of traffic. The impact is limited to voice and messaging on the affected routes rather than the platform as a whole. Details are on the Twilio status page (https://status.twilio.com/). Microsoft 365 is still closing out the outage that began August 31 and ran into September 2. Microsoft traced it to a misconfiguration in a core authentication configuration shared by several Microsoft 365 services, which stopped authentication components from deploying correctly to part of the infrastructure. Exchange Online, Outlook, Teams, SharePoint Online, OneDrive for Business, Purview, Defender XDR, Copilot, the Microsoft 365 Admin Center and Universal Print were all affected. Mailbox connectivity returned roughly 12 hours after Microsoft first acknowledged the problem, and by September 2 the company reported availability stable above 99 percent with most users no longer seeing impact. Search was the last component still recovering. A small number of Outlook reports are still trickling in today. Background: BleepingComputer (https://www.bleepingcomputer.com/news/microsoft/microsoft-exchange-online-outage-causes-email-failures-auth-issues/) and TechCrunch (https://techcrunch.com/2026/09/01/microsoft-365-outage-drags-on-but-things-are-improving/). Cloudflare is showing a minor incident: an HTTP/3 problem affecting R2 custom domains, where a fix is being implemented, and incorrect geolocation for some WARP users. Scheduled datacenter maintenance is also running in San Diego, Denver, Chicago, Montreal and Nashville. Neither issue is global (https://www.cloudflarestatus.com/). Slack is still carrying a long-running incident, open since August 13, covering trouble accessing historical messages on workspaces that have custom data retention policies enabled. Every other Slack function is reported normal. The rest of our watch list is clear. Google Cloud reports no broad severe incidents and Azure shows no active events. GitHub, Akamai, OpenAI, Anthropic, Discord, Atlassian, Zoom, Vercel, Heroku, DigitalOcean and Stripe are all operational. The AI platforms we track (Cursor, Groq, Replicate, ElevenLabs, Cohere and Stability AI) report no issues, and Meta's platforms show no active disruption. One caveat worth stating plainly: we could not load Amazon's official health dashboard directly on this run, and a single third-party aggregator flagged elevated AWS error rates. No news coverage, no substantive customer reports and no other source corroborated that, so we are treating AWS as normal while noting we could not confirm it at the source. Resolved in the past 72 hours: Discord fixed two issues on September 2, one with voice endpoint selection that could prevent calls from starting and one causing degraded performance on some servers. Anthropic cleared a 14-minute stretch of elevated errors on Claude Sonnet 5 plus a credit purchase delay, both on September 2. Cursor resolved service degradation affecting its automation and cloud agent features in early September. One thing we checked and ruled out: searches for current submarine cable damage surface articles about Red Sea cable cuts that read as recent but actually describe earlier events, including one traced to February 2024. We found no evidence of new subsea cable damage today. Overall status: incidents detected, no widespread outage. The internet is up.

No internet-wide outage today. The most active item is Twilio, which is working through a cluster of SMS delivery problems across several carrier networks. Twilio (ongoing). Twilio's status page lists multiple open SMS delivery incidents this morning: delayed delivery receipts to H3G subscribers in the United Kingdom (investigating), delivery delays across multiple networks in Spain (monitoring, recovery observed), delays from a subset of Twilio short codes to US networks (cause identified), and short-code delays to networks in Brazil. A separate issue is affecting Silent Network Auth redirects for Three UK. These are carrier-path problems affecting message delivery timing rather than a Twilio platform outage, and several are already recovering. Details: https://status.twilio.com/ Cloudflare (minor, ongoing). Two low-impact issues remain open: an HTTP/3 problem affecting R2 custom domains (since August 31, 18:47 UTC, fix being implemented) and incorrect geolocation for some WARP users (since August 27). Two short localized incidents resolved earlier today - increased latency in Taipei (08:42-09:15 UTC) and Workers issues in Narita (01:02-02:24 UTC). Details: https://www.cloudflarestatus.com/ Slack (ongoing, long-running). An incident open since August 13 continues to affect retrieval of historical messages for workspaces with custom data retention policies. Messaging and admin functions are otherwise normal. Details: https://slack-status.com/ AWS (ongoing, regional). The Middle East regions me-central-1 (UAE) and me-south-1 (Bahrain) remain degraded following physical infrastructure damage earlier this year. AWS continues to advise customers to run workloads in alternate regions and expects recovery to take months. This is long-running background, not a new event. Details: https://health.aws.amazon.com/health/status Crowd-sourced reports. Downdetector activity today points at regional carriers and single platforms rather than shared infrastructure, including Turk Telekom, Telkomsel and DeepSeek. These look like isolated provider issues. Resolved in the past 72 hours. Google Cloud recovered from network service degradation in us-central1-b that affected AlloyDB, Apigee, Filestore and about a dozen other products for 4 hours 8 minutes on September 1. GitHub resolved delays in commit processing that left pull request diffs stale on September 1. Discord resolved an incident on August 31 that made some servers, voice calls and activities unavailable. Anthropic resolved several rounds of degraded performance on claude.ai and Claude Code between August 31 and September 1, plus delayed credit purchases that cleared at 01:24 UTC today. Cursor resolved model degradation affecting Composer 2.5 and Grok requests at 03:07 UTC. Zoom resolved a Google Workspace add-on degradation on September 1. All clear elsewhere. Azure, Azure DevOps, GitHub, Akamai, OpenAI, Atlassian, Vercel, DigitalOcean, Heroku, Groq, Replicate, ElevenLabs, Cohere and Stability AI are all reporting normal operations. Stripe's status page did not render for automated retrieval this run, so its state is unconfirmed; no outage signal for Stripe appeared in any other source. Overall: incidents detected, none widespread. The internet is up.

A large Microsoft 365 outage that began on August 31 is still working through its backlog this morning. Core internet infrastructure — global routing, DNS, and the major cloud networks — is stable, and there is no internet-wide disruption. Microsoft 365 / Exchange Online — recovering. Microsoft began investigating Exchange Online failures at about 17:30 UTC on August 31, tracked as EX1464935 and later broadened to MO1465074. Users saw delayed or failed email delivery, authentication errors, and mailbox operation failures, and the impact spread to OneDrive for Business, SharePoint Online, Microsoft Teams, Microsoft Purview, and Microsoft Defender XDR. Microsoft attributed it to a misconfiguration in a core authentication configuration shared by several internal Exchange Online services, which prevented authentication components from deploying as expected to part of its infrastructure. Downdetector logged more than 5,000 reports around the peak. As of Microsoft's update early on September 1, mailbox connectivity has largely returned to normal, while search functionality and queued email backlogs are still catching up. Our August 31 report was filed before this incident began. Details: https://www.bleepingcomputer.com/news/microsoft/microsoft-exchange-online-outage-causes-email-failures-auth-issues/ Cloudflare — three minor incidents open. HTTP errors on API calls affecting D1 and Workflows, opened 08:38 UTC today and under investigation. An incorrect HTTP/3 advertisement causing delayed or failed asset loads on R2 custom domains for some Firefox users, open since August 31. And incorrect geolocation for WARP users, open since August 27 with a fix in progress. A handful of edge locations are in partial outage or maintenance, which is normal background re-routing. Source: https://www.cloudflarestatus.com OpenAI — recovering. Elevated latency on the Responses API, identified August 31. Mitigation applied, with recovery being monitored as of 01:46 UTC. Source: https://status.openai.com Twilio — degraded on carrier routes. Voice call failures between Twilio and multiple networks in Malaysia and Thailand, SMS delivery delays to H3G United Kingdom, MMS delays and failures from AT&T to Twilio toll-free numbers, and SMS delays to C-Spire, Dish, and Telefonica El Salvador. Most are showing recovery and all are rated minor. Source: https://status.twilio.com Slack — long-running incident still open. The issue that began August 13 affecting access to historical messages on workspaces with custom data retention policies remains active. Slack says custom retention processing has been re-enabled for affected active workspaces while message, edit, and file restoration continues. Source: https://slack-status.com Operating normally: AWS, Microsoft Azure, Azure DevOps, Google Cloud, GitHub, Akamai, Anthropic, Discord, Atlassian, Zoom, Vercel, Heroku, DigitalOcean, Stripe, Cursor, Groq, Replicate, ElevenLabs, Cohere, and Stability AI. Infrastructure layer: no BGP hijacks, route leaks, submarine cable faults, or DNS root-server problems were reported in the past 24 hours. Stripe has a scheduled TWINT payment-method maintenance running until 15:00 UTC today. One gap to note: Meta's status page returned no machine-readable detail on this run, so our read on Facebook, Instagram, and WhatsApp comes from third-party reporting only. We found no credible reports of a Meta outage in the last 24 hours. Recently resolved, past 72 hours: a Google Cloud us-west1 networking incident on August 20 caused by fiber maintenance is long closed, and no further cloud-provider incidents were opened between August 29 and 31 apart from those listed above. Overall status: incidents detected. One large single-vendor SaaS outage in recovery, plus several minor provider issues. The internet itself is up.