Residency is not autonomy: what Microsoft’s outage revealed

Friday 13th calendar with O365 logo

The global Microsoft 365 outage has exposed an uncomfortable distinction for Australian and New Zealand cloud customers. Keeping data within national borders is not the same as running a service that can function independently within them.

The disruption ran from 9:05pm AEDT on Thursday 12 November to 3:10am on Saturday 14 November, according to Microsoft’s incident record. It cut access to Exchange Online mailboxes and broke features across Microsoft 365 that depend on email services.

Those included Teams calendar and meeting scheduling, Copilot and Intune. Microsoft Purview was also hit, including data loss prevention, auto-labelling, audit searches and eDiscovery search and export.

In Australia, the federal government went as far as activating a formal agreement with Microsoft to secure priority in the recovery.

The outage has reopened questions about what organisations actually get when they buy cloud services carrying data residency and sovereignty assurances. For government agencies in particular, the distinction matters.

Data residency addresses where information is stored. Data sovereignty goes further, covering the laws, jurisdictions and controls that govern that information.

A third consideration is receiving much less attention: operational sovereignty. It asks whether an Australian or New Zealand cloud service can keep operating if a critical part of its provider’s global infrastructure fails.

The Microsoft outage makes that question pressing.

Microsoft’s preliminary root cause says the incident was caused by a failure in the systems responsible for establishing connections to Exchange Online. Traffic shifts during standard network maintenance and upgrades exposed a latent bug, causing a rapidly growing queue of requests.

The affected servers exhausted their processing capacity and could no longer accept new connections. Microsoft said the problem was introduced during an earlier network capacity upgrade and surfaced after a recent network change.

It described the outage as “a technical issue not related to cyber or AI / agents”.

That fits the assessment of Mahmoud Elkhodr, a senior lecturer in information and communication technologies at CQUniversity Australia. Writing in The Conversation while the outage was under way, he said the evidence did not suggest a cyberattack.

“Outages of this scale are more often self-inflicted in this way than they are the work of attackers,” he wrote.

Microsoft has promised a preliminary post-incident report within two business days and a final report within five. Neither had been published at the time of writing.

Its incident record describes the failure in terms of connection-handling systems and an “upstream authentication/routing configuration component”. It does not say where that infrastructure is located.

There is therefore no evidence that an overseas component caused the Australian and New Zealand impact. However, the available evidence also points away from a failure specific to either country.

Microsoft first reported Exchange Online access problems in the Americas, before warning that users in any region could be affected. In New Zealand, Immigration New Zealand paused processing of all visa applications.

That raises an important question for organisations buying cloud services on the strength of local data residency. Could an infrastructure dependency outside Australia or New Zealand make locally resident information inaccessible?

Residency is not autonomy

Microsoft makes significant data residency commitments for Microsoft 365. Its documentation says specified customer data at rest for eligible Australian tenants is stored within the Australian geography.

That includes Exchange Online mailbox content, SharePoint Online content, OneDrive files, Teams messages and specified Microsoft 365 Copilot content.

New Zealand customers face an extra step. Microsoft directs organisations that need residency commitments for New Zealand to its Advanced Data Residency program. Tenants with fewer than 1,645 eligible licences have no residency commitment under it.

Advanced Data Residency lists both Australia and New Zealand as Local Region Geographies. Covered services include Exchange Online, SharePoint, OneDrive, Teams, Microsoft 365 Copilot, Defender for Office 365, Office for the web and selected Purview services.

These are meaningful assurances. But they principally answer one question: where is the customer’s information?

They do not, by themselves, answer another: where are all the systems that access to that information depends on?

A hyperscale cloud is not simply a collection of servers holding customer files. Services depend on networking, authentication, monitoring, orchestration, security and traffic management.

This outage illustrates the point. Microsoft’s preliminary explanation places the failure in the systems that establish connections to the service, not in the storage holding customer data.

The incident record offers further clues. During recovery, Microsoft redistributed connection traffic across its worldwide infrastructure.

At one point it said it was evaluating whether isolating traffic within each region might allow affected regions to recover independently. That indicates regional recovery was not automatic, and that isolating regions was an option Microsoft had to evaluate mid-incident.

Knowing that data sits in an Australian or New Zealand facility may therefore be only part of the sovereignty risk picture.

A place in the recovery queue

Australia did not simply wait for Microsoft to restore service. On Friday 13 November, Home Affairs Minister Tony Burke said the government had activated its digital resilience agreement with Microsoft.

The memorandum of understanding, announced in June, is the first of the government’s digital resilience agreements with technology providers. One of its five commitments deals directly with outages.

Under it, the parties agree to “work together to prioritise the recovery of Australia’s digital economy as soon as possible” from significant technology outages in the Microsoft ecosystem.

Microsoft’s incident record shows recovery actions completed across parts of its Asia Pacific infrastructure at 10:01pm AEDT on 13 November. It reported comparable progress in Europe and North America later that night.

The memorandum is explicitly non-binding. It also says it is not intended to confer exclusivity or preference, or to limit the Commonwealth’s engagement with other technology providers.

Much about it remains undisclosed. Non-public information exchanged under the agreement is confidential, and public disclosure is limited to its existence and broad objectives.

So it is not yet clear what activation changed in practice, or which Australian organisations benefited.

For the sovereignty debate, the agreement cuts both ways. It shows the Australian Government recognises that access to locally stored data depends on decisions made inside a global provider.

But the remedy is diplomatic rather than architectural. The agreement does not make Australian infrastructure independent. It seeks a place near the front of a global recovery queue, on non-binding terms.

No equivalent New Zealand arrangement has been made public.

The PROTECTED question

For Australian government, the issue is especially significant. Microsoft says Azure and Office 365 have been assessed under the Infosec Registered Assessors Program (IRAP) for handling information at the PROTECTED classification.

IRAP assessments examine cloud security controls against Australian Government requirements. Assessors identify mitigations and residual risks.

But the sovereignty question raised by a global outage is subtly different.

Suppose customer data remains securely stored in Australia, but access to it depends on infrastructure or a control-plane service operating elsewhere.

If that overseas dependency fails, Australian government users may be unable to reach an intact Australian dataset. No data needs to leave Australia, and no residency commitment needs to be breached.

Yet the organisation has still lost sovereign access to its information. That is the distinction between data residency and operational sovereignty.

The scenario remains hypothetical for this incident until Microsoft publishes more architectural detail. If its post-incident report shows the initiating failure occurred in the US or elsewhere offshore, the outage will be a powerful real-world demonstration of that distinction.

That would not necessarily mean Microsoft had failed its residency commitments. It would expose a different sovereign risk: locally resident data that cannot be reached because of a failure beyond the country’s borders.

Sovereignty is of limited operational comfort if the data stays safely at home but the systems needed to reach it do not.

New Zealand faces the same question. Its government cloud policy requires agencies to consider public cloud services case by case, following a risk assessment. Its Cloud Jurisdictional Risk guidance addresses the jurisdictional risks associated with cloud providers.

The New Zealand National Cyber Security Centre’s Cyber Threat Report 2026 names digital supply chains as a major risk to New Zealand businesses. It notes that many organisations now rely on a mix of platforms, suppliers and outsourced services, many of them based overseas.

The report warns that a single security failure at vendor level, here or overseas, can cascade. It cites the attack on Canvas, a US-owned, cloud-hosted learning management system, after which the US-based operator took several days to restore services.

That observation applies beyond security breaches. The architectural concentration that creates cyber supply chain risk also creates availability risk. A service may be locally hosted while remaining globally dependent.

Four questions for buyers

For government CIOs, CISOs and information managers, cloud sovereignty may need to be considered in layers.

Data residency: Where is our information physically stored?

Data sovereignty: Which jurisdictions, laws and administrative controls can apply to it?

Operational sovereignty: Which infrastructure, services and jurisdictions must remain operational for us to access and use it?

Recovery priority: When a global outage hits, where do we sit in the queue, and what commits the provider to that?

These questions grow more important as organisations consolidate identity, communications, records, collaboration, security and AI onto a handful of hyperscale platforms.

An organisation can satisfy the first requirement while retaining dependencies outside national borders under the third. That does not automatically make the architecture insecure or inappropriate. It does mean the dependency should be understood.

Microsoft’s post-incident reports will test how transparent hyperscalers are prepared to be. They could establish what failed, where it was located and why its failure reached customers across multiple regions.

For Australian and New Zealand buyers, a further question follows. Could the affected regional infrastructure operate independently, or did it need elements of Microsoft’s wider global platform?

That cannot be answered from public information yet. But perhaps it should become a standard procurement question.

Governments have spent years asking cloud providers where their data resides. The Microsoft outage suggests the next question may be just as important.

What has to keep working outside our country for us to access the data that remains inside it?

Friday 13th calendar with O365 logo

The global Microsoft 365 outage has exposed an uncomfortable distinction for Australian and New Zealand cloud customers. Keeping data within national borders is not the same as running a service that can function independently within them.

The disruption ran from 9:05pm AEDT on Thursday 12 November to 3:10am on Saturday 14 November, according to Microsoft’s incident record. It cut access to Exchange Online mailboxes and broke features across Microsoft 365 that depend on email services.

Those included Teams calendar and meeting scheduling, Copilot and Intune. Microsoft Purview was also hit, including data loss prevention, auto-labelling, audit searches and eDiscovery search and export.

In Australia, the federal government went as far as activating a formal agreement with Microsoft to secure priority in the recovery.

The outage has reopened questions about what organisations actually get when they buy cloud services carrying data residency and sovereignty assurances. For government agencies in particular, the distinction matters.

Data residency addresses where information is stored. Data sovereignty goes further, covering the laws, jurisdictions and controls that govern that information.

A third consideration is receiving much less attention: operational sovereignty. It asks whether an Australian or New Zealand cloud service can keep operating if a critical part of its provider’s global infrastructure fails.

The Microsoft outage makes that question pressing.

Microsoft’s preliminary root cause says the incident was caused by a failure in the systems responsible for establishing connections to Exchange Online. Traffic shifts during standard network maintenance and upgrades exposed a latent bug, causing a rapidly growing queue of requests.

The affected servers exhausted their processing capacity and could no longer accept new connections. Microsoft said the problem was introduced during an earlier network capacity upgrade and surfaced after a recent network change.

It described the outage as “a technical issue not related to cyber or AI / agents”.

That fits the assessment of Mahmoud Elkhodr, a senior lecturer in information and communication technologies at CQUniversity Australia. Writing in The Conversation while the outage was under way, he said the evidence did not suggest a cyberattack.

“Outages of this scale are more often self-inflicted in this way than they are the work of attackers,” he wrote.

Microsoft has promised a preliminary post-incident report within two business days and a final report within five. Neither had been published at the time of writing.

Its incident record describes the failure in terms of connection-handling systems and an “upstream authentication/routing configuration component”. It does not say where that infrastructure is located.

There is therefore no evidence that an overseas component caused the Australian and New Zealand impact. However, the available evidence also points away from a failure specific to either country.

Microsoft first reported Exchange Online access problems in the Americas, before warning that users in any region could be affected. In New Zealand, Immigration New Zealand paused processing of all visa applications.

That raises an important question for organisations buying cloud services on the strength of local data residency. Could an infrastructure dependency outside Australia or New Zealand make locally resident information inaccessible?

Residency is not autonomy

Microsoft makes significant data residency commitments for Microsoft 365. Its documentation says specified customer data at rest for eligible Australian tenants is stored within the Australian geography.

That includes Exchange Online mailbox content, SharePoint Online content, OneDrive files, Teams messages and specified Microsoft 365 Copilot content.

New Zealand customers face an extra step. Microsoft directs organisations that need residency commitments for New Zealand to its Advanced Data Residency program. Tenants with fewer than 1,645 eligible licences have no residency commitment under it.

Advanced Data Residency lists both Australia and New Zealand as Local Region Geographies. Covered services include Exchange Online, SharePoint, OneDrive, Teams, Microsoft 365 Copilot, Defender for Office 365, Office for the web and selected Purview services.

These are meaningful assurances. But they principally answer one question: where is the customer’s information?

They do not, by themselves, answer another: where are all the systems that access to that information depends on?

A hyperscale cloud is not simply a collection of servers holding customer files. Services depend on networking, authentication, monitoring, orchestration, security and traffic management.

This outage illustrates the point. Microsoft’s preliminary explanation places the failure in the systems that establish connections to the service, not in the storage holding customer data.

The incident record offers further clues. During recovery, Microsoft redistributed connection traffic across its worldwide infrastructure.

At one point it said it was evaluating whether isolating traffic within each region might allow affected regions to recover independently. That indicates regional recovery was not automatic, and that isolating regions was an option Microsoft had to evaluate mid-incident.

Knowing that data sits in an Australian or New Zealand facility may therefore be only part of the sovereignty risk picture.

A place in the recovery queue

Australia did not simply wait for Microsoft to restore service. On Friday 13 November, Home Affairs Minister Tony Burke said the government had activated its digital resilience agreement with Microsoft.

The memorandum of understanding, announced in June, is the first of the government’s digital resilience agreements with technology providers. One of its five commitments deals directly with outages.

Under it, the parties agree to “work together to prioritise the recovery of Australia’s digital economy as soon as possible” from significant technology outages in the Microsoft ecosystem.

Microsoft’s incident record shows recovery actions completed across parts of its Asia Pacific infrastructure at 10:01pm AEDT on 13 November. It reported comparable progress in Europe and North America later that night.

The memorandum is explicitly non-binding. It also says it is not intended to confer exclusivity or preference, or to limit the Commonwealth’s engagement with other technology providers.

Much about it remains undisclosed. Non-public information exchanged under the agreement is confidential, and public disclosure is limited to its existence and broad objectives.

So it is not yet clear what activation changed in practice, or which Australian organisations benefited.

For the sovereignty debate, the agreement cuts both ways. It shows the Australian Government recognises that access to locally stored data depends on decisions made inside a global provider.

But the remedy is diplomatic rather than architectural. The agreement does not make Australian infrastructure independent. It seeks a place near the front of a global recovery queue, on non-binding terms.

No equivalent New Zealand arrangement has been made public.

The PROTECTED question

For Australian government, the issue is especially significant. Microsoft says Azure and Office 365 have been assessed under the Infosec Registered Assessors Program (IRAP) for handling information at the PROTECTED classification.

IRAP assessments examine cloud security controls against Australian Government requirements. Assessors identify mitigations and residual risks.

But the sovereignty question raised by a global outage is subtly different.

Suppose customer data remains securely stored in Australia, but access to it depends on infrastructure or a control-plane service operating elsewhere.

If that overseas dependency fails, Australian government users may be unable to reach an intact Australian dataset. No data needs to leave Australia, and no residency commitment needs to be breached.

Yet the organisation has still lost sovereign access to its information. That is the distinction between data residency and operational sovereignty.

The scenario remains hypothetical for this incident until Microsoft publishes more architectural detail. If its post-incident report shows the initiating failure occurred in the US or elsewhere offshore, the outage will be a powerful real-world demonstration of that distinction.

That would not necessarily mean Microsoft had failed its residency commitments. It would expose a different sovereign risk: locally resident data that cannot be reached because of a failure beyond the country’s borders.

Sovereignty is of limited operational comfort if the data stays safely at home but the systems needed to reach it do not.

New Zealand faces the same question. Its government cloud policy requires agencies to consider public cloud services case by case, following a risk assessment. Its Cloud Jurisdictional Risk guidance addresses the jurisdictional risks associated with cloud providers.

The New Zealand National Cyber Security Centre’s Cyber Threat Report 2026 names digital supply chains as a major risk to New Zealand businesses. It notes that many organisations now rely on a mix of platforms, suppliers and outsourced services, many of them based overseas.

The report warns that a single security failure at vendor level, here or overseas, can cascade. It cites the attack on Canvas, a US-owned, cloud-hosted learning management system, after which the US-based operator took several days to restore services.

That observation applies beyond security breaches. The architectural concentration that creates cyber supply chain risk also creates availability risk. A service may be locally hosted while remaining globally dependent.

Four questions for buyers

For government CIOs, CISOs and information managers, cloud sovereignty may need to be considered in layers.

Data residency: Where is our information physically stored?

Data sovereignty: Which jurisdictions, laws and administrative controls can apply to it?

Operational sovereignty: Which infrastructure, services and jurisdictions must remain operational for us to access and use it?

Recovery priority: When a global outage hits, where do we sit in the queue, and what commits the provider to that?

These questions grow more important as organisations consolidate identity, communications, records, collaboration, security and AI onto a handful of hyperscale platforms.

An organisation can satisfy the first requirement while retaining dependencies outside national borders under the third. That does not automatically make the architecture insecure or inappropriate. It does mean the dependency should be understood.

Microsoft’s post-incident reports will test how transparent hyperscalers are prepared to be. They could establish what failed, where it was located and why its failure reached customers across multiple regions.

For Australian and New Zealand buyers, a further question follows. Could the affected regional infrastructure operate independently, or did it need elements of Microsoft’s wider global platform?

That cannot be answered from public information yet. But perhaps it should become a standard procurement question.

Governments have spent years asking cloud providers where their data resides. The Microsoft outage suggests the next question may be just as important.

What has to keep working outside our country for us to access the data that remains inside it?