Interfaz en español. Las fichas, los resúmenes y las fuentes conservan su idioma original.
Microsoft 365 outage disrupts services and AI connectors
Análisis del incidente
Microsoft reports a disruption to several Microsoft 365 services, with recovery still in progress at the review cutoff. OpenAI and Anthropic separately confirm Microsoft connector failures, including impact to Claude for Outlook.
Quiénes deben revisar: Microsoft 365 administrators, operations teams and teams using Microsoft connectors in ChatGPT, Claude or automated workflows.
Check provider status and identify affected workflows. Hold repeated writes when outcomes are unknown; verify sent messages, calendar changes and pending jobs before retrying.
¿Qué ocurrió y por qué es relevante?
At the November 12 review cutoff, Microsoft still reported service degradation. Its 20:32 UTC update describes failed health checks that removed servers from service and investigation of elevated processor usage. A targeted change was being tested on a subset of servers; this was not a recovery confirmation.
The potentially affected services include Exchange Online, Teams, SharePoint Online, Microsoft 365 Copilot, Defender XDR, Planner and To Do. The public page also warns that administrators may be unable to reach the admin center.
OpenAI’s 18:29 UTC incident update reports problems creating Microsoft connections and using connector tools. Anthropic’s 20:28 UTC update says most mail, calendar and search requests through its Microsoft 365 connector are failing. Claude for Outlook is also affected, including mail/calendar actions and opening the add-in for some users. Earlier Anthropic guidance also covered SharePoint and Teams chat requests.
These notices establish availability problems. They do not confirm an attack, data theft or a connection to the separate Outlook vulnerability. They do not establish that every customer or all OpenAI and Anthropic services are affected.
Oktsec recommends identifying dependent workflows, preserving uncertain job results and checking what actually completed before retries. Pause agent actions that could duplicate messages or calendar changes, and verify each integration before processing its backlog.
- Proveedor / autor
- Microsoft
- Divulgado
- 12 nov 2026
- Última revisión de evidencia
- 12 nov 2026 UTC
¿Qué sistemas y versiones están afectados?
| Producto | Afectado | Versión corregida / acción |
|---|---|---|
| Microsoft 365 business and enterprise services | Microsoft lists potential impact to Exchange Online, Teams, SharePoint Online, Microsoft 365 Copilot, Defender XDR, Planner and To Do; administrators may also have trouble reaching the admin center. | Provider recovery in progress at the review cutoff; no customer software patch is identified in this status notice. |
| Microsoft connectors in ChatGPT | OpenAI reports that new connections may fail and connector tools may not work as expected. This notice does not establish an outage of all ChatGPT or API functionality. | OpenAI is investigating; validate connector operations before resuming dependent workflows. |
| Microsoft 365 connector in Claude and Claude for Outlook | Anthropic reports failures for most mail, calendar and search requests; new connections may not complete. Claude for Outlook mail/calendar actions and opening the add-in may also fail. | Anthropic is working with its upstream provider; validate operations after recovery is announced. |
Usá las indicaciones para tu producto y rama. Confirmá con el proveedor las correcciones adaptadas a tu distribución; el nombre del producto por sí solo no establece exposición.
¿Qué evidencia respalda este aviso?
Evidencia del incidente
Fundamento de la prioridad
An ongoing availability incident can interrupt business workflows and leave connector actions incomplete or uncertain. Immediate attention concerns continuity and safe retries, not confirmed exploitation.
¿Qué implica para los entornos de IA?
A connector error can leave an action’s outcome unknown. Oktsec recommends pausing dependent agent writes, retaining job identifiers and reconciling results before retries so failed reads do not produce misleading conclusions and repeated writes do not duplicate work.
Fuentes de este aviso
- Anthropic: Microsoft 365 connector and Claude for OutlookAnthropic · Revisado 12 nov 2026 UTC
- OpenAI: Microsoft connector errorsOpenAI · Revisado 12 nov 2026 UTC
- Microsoft: multiple services degraded; update at 20:32 UTCMicrosoft · Revisado 12 nov 2026 UTC
- Microsoft public status detailsMicrosoft · Revisado 12 nov 2026 UTC
Historial editorial1 entrada
Added the confirmed availability incident, separate connector impacts and operational guidance with an explicit source cutoff.
Prepará tu respuesta
Usá esta lista para evaluar si el aviso afecta a tus sistemas, registrar las acciones de tu equipo y compartir el resultado. Radar no escanea ni verifica tu entorno.
Los cambios se guardan automáticamente en este navegador. No se sincronizan con tu equipo ni se envían a Oktsec.
1. Evaluá si aplica a tu entorno
Revisá los sistemas y versiones afectados y registrá tu decisión a continuación. Si tenés dudas, elegí «Requiere verificación».
2. Registrá la respuesta
Indicá el sistema y la persona o equipo responsable. Actualizá cada acción y agregá una referencia al ticket o notas sobre lo revisado.
Identify dependent services and workflows
Check Microsoft’s public status page and the OpenAI or Anthropic incident relevant to each connector. Oktsec recommends recording which mail, calendar, search and automation workflows are actually failing in your environment.
Agregar evidencia o notas
Usá solo referencias. No guardes credenciales ni datos confidenciales acá.Preserve operations with an unknown result
Oktsec recommends retaining timestamps, job identifiers and errors. A timeout does not prove that a message was not sent or a calendar change was not applied; check the authoritative result before repeating the action.
Agregar evidencia o notas
Usá solo referencias. No guardes credenciales ni datos confidenciales acá.Pause repeated writes and use agreed alternatives
Oktsec recommends holding automatic retries for actions that could duplicate work and using established continuity channels for urgent tasks. Avoid repeatedly reconnecting accounts or changing credentials solely because of this outage.
Agregar evidencia o notas
Usá solo referencias. No guardes credenciales ni datos confidenciales acá.Verify recovery before clearing the backlog
After the relevant provider reports recovery, Oktsec recommends checking reads and writes through each affected integration, reconciling uncertain operations and resuming queued work gradually. Recovery of one service does not verify all connectors.
Agregar evidencia o notas
Usá solo referencias. No guardes credenciales ni datos confidenciales acá.
3. Compartí el plan
Descargá un archivo Markdown con tu decisión, los estados de las acciones, las notas y las fuentes. Compartilo con tu equipo o adjuntalo a un ticket.
Sistemas afectados y correcciones
Compará la versión en ejecución y la configuración con las indicaciones del proveedor.
| Producto | Afectado | Versión corregida / acción |
|---|---|---|
| Microsoft 365 business and enterprise services | Microsoft lists potential impact to Exchange Online, Teams, SharePoint Online, Microsoft 365 Copilot, Defender XDR, Planner and To Do; administrators may also have trouble reaching the admin center. | Provider recovery in progress at the review cutoff; no customer software patch is identified in this status notice. |
| Microsoft connectors in ChatGPT | OpenAI reports that new connections may fail and connector tools may not work as expected. This notice does not establish an outage of all ChatGPT or API functionality. | OpenAI is investigating; validate connector operations before resuming dependent workflows. |
| Microsoft 365 connector in Claude and Claude for Outlook | Anthropic reports failures for most mail, calendar and search requests; new connections may not complete. Claude for Outlook mail/calendar actions and opening the add-in may also fail. | Anthropic is working with its upstream provider; validate operations after recovery is announced. |
Confirmá con el proveedor las correcciones adaptadas a tu distribución. Coincidir en el producto no basta para establecer exposición.
Evidencia y fuentes
Por qué tiene esta prioridad
An ongoing availability incident can interrupt business workflows and leave connector actions incomplete or uncertain. Immediate attention concerns continuity and safe retries, not confirmed exploitation.
Criterios de prioridadFuentes
Los datos técnicos provienen de estas fuentes. Las prioridades y las acciones recomendadas son el análisis de Oktsec.
- Anthropic: Microsoft 365 connector and Claude for OutlookAnthropic · Aviso del proveedor · Revisado 12 nov 2026 UTC
- OpenAI: Microsoft connector errorsOpenAI · Aviso del proveedor · Revisado 12 nov 2026 UTC
- Microsoft: multiple services degraded; update at 20:32 UTCMicrosoft · Aviso del proveedor · Revisado 12 nov 2026 UTC
- Microsoft public status detailsMicrosoft · Aviso del proveedor · Revisado 12 nov 2026 UTC
Registro técnicoProductos, identificadores y detalles de respaldo
| Producto / servicio | Microsoft 365, Microsoft connectors in ChatGPT, Microsoft 365 connector in ClaudeMicrosoft |
|---|---|
| Evidencia | Incidente confirmado por el proveedor |
| Exposure | Depende de la aplicación o configuración |
- Divulgado
- 12 nov 2026
- Registro revisado
- 12 nov 2026 UTC
Para entornos con agentes de IA
A connector error can leave an action’s outcome unknown. Oktsec recommends pausing dependent agent writes, retaining job identifiers and reconciling results before retries so failed reads do not produce misleading conclusions and repeated writes do not duplicate work.