Join our Newsletter — 33% off our NHI Course
Home› Guides› Shared Responsibility vs Shared Fate: What Changes for…
Guide Cloud Security

Shared Responsibility vs Shared Fate: What Changes for Cloud Customers

PDF for newsletter subscribers · free to read online

← All guides
By NHI Mgmt Group, edited by Lalit Choda Updated 14 November 2026 17 min read
On this page

Shared fate is often read as the cloud provider taking on part of the customer's risk, but it changes how much help the customer gets, not who is accountable for the customer's half. Shared responsibility draws a line between what the provider secures and what the customer secures. Shared fate, a term Google Cloud uses for its own approach, keeps that line and adds provider effort on the customer's side of it: secure defaults, deployable blueprints, compliance guardrails and a route to cyber insurance. This guide compares the two models as concepts, explains what shared fate does and does not change in contracts and audits, and sets out what customers still own whichever label their provider uses. It is written for cloud security architects, GRC and procurement leads, and anyone negotiating or relying on a cloud agreement. The baseline model itself is covered in the Shared Responsibility Models Guide; this page goes further on shared fate.

Where providers are named, they are listed alphabetically: AWS, Google Cloud, Microsoft. Order is not a ranking.

Key takeaways

  • Shared fate builds on shared responsibility, it does not replace it. Google describes shared fate as building on the shared responsibility model, and states that customers always remain responsible for their access policies and data.
  • Defaults move work, not liability. A provider that blocks public storage or service account keys by default removes a common mistake, but the customer can still switch the default off, and the outcome is then the customer's.
  • Insurance tie-ins are underwriting, not cover from the provider. Google's Risk Protection Program lets customers share a posture report with partner insurers; Google states it is not a licensed broker and does not issue policies.
  • Contracts rarely change. Shared fate tooling sits in documentation and product settings; the binding allocation of duties is still the customer agreement, the service terms and any negotiated security schedule, so check those, not the marketing model.
  • Treat provider defaults as a baseline you verify. Inventory which defaults apply to which resources and when they started, enforce them yourself with policy, and measure drift, because older accounts and resources often predate them.

The short answer

Shared responsibility answers the question "whose job is this control?" Shared fate answers a different question: "how much will the provider help me do my part well?" Under both, the provider runs the physical infrastructure and the services it operates, and the customer owns its data, its identities and access policies, and how it configures the services it uses. Shared fate adds secure-by-default settings, reference architectures deployed as code, compliance boundaries and insurance pathways, which lower the chance that the customer's half fails. It does not move accountability for that half to the provider.

So the practical choice is not between the two models. It is how far to adopt the provider's opinionated defaults and tooling, and how to record in your own control set and contracts which defaults you rely on.

Side-by-side comparison

DimensionShared responsibilityShared fate
Core ideaA division of tasks: provider secures the cloud, customer secures what it puts in itThe same division, plus active provider help on the customer's side, described by Google as an ongoing partnership to improve security
Where the line sitsMoves with the service model: most customer work in IaaS, least in SaaSSame line; data and access policies stay with the customer in every model
Provider deliverablesSecure infrastructure, compliance reports, configurable security featuresAlso secure defaults, blueprints with recommendations enabled, compliance guardrails, posture tooling, insurance pathway
Default postureFeatures available, customer chooses whether to enableSelected protections on by default for new resources or organisations, customer may change them
Main risk addressedUnclear ownershipMisconfiguration in the customer's half
Evidence for auditorsProvider SOC 2 and ISO/IEC 27001 reports for inherited controls; customer evidence for its ownSame, plus inherited defaults such as default encryption that Google says can serve as evidence of posture
Contract effectDefined in the customer agreement and service termsUsually none by itself: tooling and guidance sit outside the core agreement unless negotiated in

Each model in depth

Shared responsibility

Threat model: the attacker targets whatever the customer has left open, because the provider's layers are hardened at scale. The realistic threats are stolen credentials, over-permissive identities, exposed storage, unpatched guest operating systems and missing logging. Architecture in words: the provider publishes a matrix per service model, the customer maps each service it uses to that matrix, names owners for its rows and enforces them with its own guardrails. Failure modes: unowned shared rows (patching, configuration, training), SaaS treated as provider-secured, and evidence collected for the provider's half but not the customer's. Controls: a per-service responsibility matrix, organisation-level preventive policy, identity governance, logging and posture management. The Shared Responsibility Matrix Template gives a ready structure.

Worked scenario: a retailer moves its order database to a managed database service. The provider patches the engine and hosts; the retailer still decides network exposure, credentials, MFA and logging. When an analyst's reused password is stolen by infostealer malware, nothing in the provider's half stops the login, because the access policy was the customer's. This is the pattern in the Snowflake breach, where attackers used stolen credentials against customer accounts that had no MFA and no network allow lists.

Shared fate

Threat model: the same attackers, with an added recognition that customers misconfigure. Google's framework page states that many cloud security breaches are the direct result of misconfiguration and that products change constantly. Shared fate aims to shrink the number of insecure states a new customer can reach by accident. Architecture in words: the customer starts from a provider blueprint deployed as infrastructure as code, with the provider's recommendations enabled by default; organisation-level constraints block risky actions; compliance packages such as Assured Workloads apply guardrails to folders holding regulated workloads; posture tooling reports drift; and the posture report can be shared with insurers. Failure modes: older organisations that predate the defaults, blueprints forked and never updated, defaults disabled to unblock a project with no compensating control, and an insurance quote read as evidence of security. Controls: keep the provider defaults under your own policy as code, track exceptions, and re-baseline when the provider changes its defaults.

Worked scenario: a fintech opens a new Google Cloud organisation. Service account key creation is blocked by default, so a developer who wants a JSON key for a CI pipeline is pushed towards workload identity federation instead. Six months later a platform team deletes the constraint to integrate a legacy tool. The provider did its part by starting secure; the decision to remove the constraint, and the long-lived key that follows, sit in the customer's half. The fix is an exception process that scopes the change to one project, records an owner and an expiry, and alerts if keys appear elsewhere. See the Cloud Workload Identity Guide.

What providers now do by default

Shared fate is Google's name, but secure by default changes reach across providers. CISA's Secure by Design programme argues that the cyber security burden falls disproportionately on consumers and small organisations, and defines secure by default as products that are secure out of the box with little or no configuration and at no extra cost. Its Secure by Design Pledge, voluntary and not legally binding, includes goals on MFA, default passwords and evidence of intrusions. Microsoft's Secure Future Initiative, launched in November 2023, states the principle that protections are enabled and enforced by default and are not optional. The table lists examples each provider documents.

ProviderDefault or provider-led controlWhat it doesWhat the customer still decides
AWSAmazon S3 new bucket settingsAll four Block Public Access settings are enabled by default, Object Ownership defaults to Bucket owner enforced with ACLs disabled, and SSE-S3 is the base level of encryption for every bucketBucket policies, access point policies, whether to turn Block Public Access settings off, and whether to use SSE-KMS keys it controls
AWSEC2 account-level IMDS defaultAn account can set IMDSv2 as the default so new instances launch with tokens required; enforcement can block launches that are not IMDSv2-onlyWhether to set the default and enforcement, per Region; existing instances are not reset
Google CloudSecure-by-default organisation policiesFor organisations created on or after 3 May 2024, baseline constraints are enforced, including iam.managed.disableServiceAccountKeyCreation, iam.disableServiceAccountKeyUpload, iam.automaticIamGrantsForDefaultServiceAccounts, iam.allowedPolicyMemberDomains and storage.uniformBucketLevelAccessWhether to keep, scope or delete each constraint (requires the Organization Policy Administrator role)
Google CloudEnterprise foundations blueprint and Assured WorkloadsBlueprints ship with security recommendations enabled by default; Assured Workloads applies predefined control packages (for example FedRAMP Moderate and High, IL4, IL5, CJIS, EU Data Boundary) to foldersWhich blueprint version to run, how to maintain the fork, which workloads go in regulated folders
MicrosoftMandatory MFA for Azure and admin portalsPhase 1 requires MFA for Azure portal, Microsoft Entra admin center and Intune admin center operations; Phase 2, from 1 October 2025, extends it to Azure CLI, Azure PowerShell, IaC tools and REST API create, update and delete operationsWhether to postpone (Phase 2 postponement was offered to 1 July 2026, with Microsoft noting this adds risk), and how to secure workload identities, which these phases do not cover

Each default applies from a date or to new resources only. That is the single most important point for a customer: the default describes the provider's starting point, not the state of your estate. Defaults and dates are as documented in November 2026. Provider hardening detail is in the AWS Security Guide, the Google Cloud Security Guide and the Microsoft Azure Security Guide.

Insurance tie-ins: what they are and are not

Google's Risk Protection Program (marked Preview as of November 2026) connects Google Cloud customers with partner cyber insurers. Its Cyber Insurance Hub, powered by Security Command Center, produces organisation-level posture reports aligned to the CIS Google Cloud Platform Foundation Benchmark, and the customer decides whether and when to send a report to the partners. As of November 2026 Google's product page names Beazley, Chubb and Munich Re as partners, states that pricing is set by the insurers, that a licensed broker is needed, that the hub only scans Google Cloud workloads, and that Google is not a licensed insurance broker and does not issue policies.

  • What it changes: underwriting can use measured posture rather than a questionnaire, and a better posture may support better terms.
  • What it does not change: the provider does not insure you, and the policy terms, exclusions and conditions are between you and the insurer. A report covers Google Cloud only, not your SaaS estate, endpoints or other clouds.
  • What to check: whether policy conditions require you to maintain the reported posture, whether a drop in posture after binding affects a claim, and how the cover interacts with any existing cyber policy. See the Cyber Insurance Guide and the Cyber Insurance Claims Process Guide.

What shared fate changes contractually

Very little, unless you negotiate it. Responsibility models, blueprints and framework pages are guidance, and Google's shared fate page itself contains no contractual terms. The binding documents remain the provider's customer agreement, service-specific terms, data processing terms, service level agreements and any security addendum. Shared fate gives you better material to negotiate with.

Contract areaUnder a plain shared responsibility readingWhat to ask for when relying on shared fate features
Security responsibilitiesProvider model referenced or absentA responsibility schedule per service you use, naming which provider defaults you rely on as controls
Change to defaultsProvider may change features and defaultsAdvance notice of changes to security defaults and blueprints that affect relied-on controls
Incident managementNotification duties for provider-side incidentsDefined notification times, cooperation in customer-side incidents and log access; Google itself calls incident management an often overlooked area where responsibilities are not clearly defined
Evidence and auditProvider third-party reportsAccess to current SOC 2 and ISO/IEC 27001 reports, the complementary user entity controls, and compliance package documentation such as Assured Workloads boundaries
Data location and personnel accessRegion choice by customerWritten commitments on data boundaries and support personnel location where compliance packages are used
InsuranceNoneClarity that the provider is not party to any insurance policy and what data is shared with insurers, and when

CSF 2.0 subcategory GV.SC-05 expects supply chain security requirements to be written into contracts and other agreements, and NIST SP 800-53 SA-9 External System Services and SA-4 Acquisition Process are the control hooks. For clause wording see the Third-Party Security Contract Clauses Template, and for deciding when provider reports are enough see Right to Audit vs Assurance Reports. Regional frameworks can also set expectations: see the Singapore Shared Responsibility Framework Guide.

What customers still own

Every provider model, including shared fate, leaves these with the customer:

  • Access policies and identities: who and what can do what, including service accounts, roles, keys, OAuth apps and agents. Microsoft's mandatory MFA applies to user sign-ins, and Microsoft states that managed identities and service principals are outside both phases, so workload identity security stays a customer design task. See the Cloud IAM Review Checklist.
  • Data: classification, location, retention and encryption key choices. Google notes, for example, that customers with EU residency needs are responsible for keeping data in EU regions.
  • Configuration after day one: blueprints start you secure; every later change is yours. See the Cloud Misconfiguration Guide.
  • Choosing controls: Google's page puts identifying required controls, from regulation, internal standards and customer requirements, on the customer.
  • Detection and response in your tenancy: enabling logs, reviewing them and responding. See the Cloud Incident Response Guide.
  • Exit and concentration risk: deeper adoption of one provider's blueprints and packages increases dependency. See the Cloud Concentration Risk and Exit Strategy Guide.

When to lean on each

SituationLean towardsWhy
Small team, new cloud estate, one providerProvider defaults and blueprints adopted largely as shippedFastest route to a sound baseline with little engineering effort
Large or long-lived estate with accounts older than the defaultsYour own responsibility matrix and policy as code, using provider defaults as referenceDefaults apply to new resources only; you must back-fill the rest
Regulated workloads (public sector, defence, criminal justice, EU sovereignty)Provider compliance packages plus your own control mappingPackages enforce boundaries, but the auditor still tests your controls
Multi-cloudOne customer-owned control baseline mapped to each provider's defaultsDefaults differ per provider; a single baseline avoids gaps between them
Seeking cyber insurancePosture reporting tied to a benchmark, kept currentUnderwriters price measured posture; drift after binding is your risk

Implementing a shared fate aware control set

  1. Inventory defaults per provider: list each documented default and its start date or scope (new organisations, new buckets, account-level opt-in), and record which of your accounts, organisations and subscriptions predate it.
  2. Map defaults to your controls: in the responsibility matrix, mark which customer controls are met by a provider default, and keep the customer as owner of each row.
  3. Codify the defaults yourself: express them as your own organisation policies, SCPs or Azure Policy assignments so they apply to old and new resources alike. Run new policies in audit or dry-run mode first, then enforce in a pilot folder or account, then estate-wide.
  4. Define the exception route: scoped, owned, time-limited exceptions with a compensating control, and a documented rollback (detach or scope down the policy) if enforcement breaks production.
  5. Adopt blueprints deliberately: pin a version, record local changes, and schedule a review against upstream releases.
  6. Negotiate the contract schedule: notice of default changes, incident cooperation, evidence access and data boundary commitments.
  7. Connect posture to insurance and risk reporting: if you use an insurer posture report, schedule it and alert on regressions.

Detection and measurement

Monitor the events that show a default being removed or bypassed: organisation policy deletions and changes in Google Cloud audit logs, S3 Block Public Access and bucket policy changes and EC2 metadata option changes in AWS CloudTrail, and Azure Policy assignment and exemption changes in the Azure Activity log. See the Cloud Logging and Audit Trail Guide.

  • Default coverage: percentage of accounts, projects and subscriptions where each relied-on default is enforced by your own policy.
  • Exception count and age: open exceptions to baseline constraints, and the share past expiry.
  • Pre-default estate: number of organisations, accounts and resources created before a default took effect and not yet back-filled.
  • Workload identity credentials: count of long-lived service account keys and access keys, trending down.
  • Posture score drift: change in benchmark-aligned posture between scheduled reports.

Common failure patterns

  • Reading shared fate as shared liability: procurement records the provider as owner of a control because a default exists. Ownership stays with the customer; the default is a means.
  • Assuming defaults are retroactive: new-organisation constraints and new-bucket settings do not reach older estates, which is where most exposed resources live.
  • Removing a default to unblock a project: a constraint deleted at organisation level for one integration opens every project. Scope exceptions narrowly.
  • Option left off: protections offered as settings protect nothing until enabled. In the Snowflake incident MFA and network policies existed as options but were not enabled on the affected accounts, and in the Capital One breach a role reachable through the metadata service could read far more data than it needed.
  • Insurance quote as assurance: a posture report is a point-in-time scan of one provider, not an audit of your security programme.

Framework mapping

ActivityNIST CSF 2.0NIST SP 800-53 Rev. 5CIS Controls v8.1ISO/IEC 27001:2022 Annex A
Allocate and record responsibilities with providersGV.RR-02; GV.SC-05SA-9 External System ServicesControl 15 Service Provider ManagementA.5.19, A.5.20
Prioritise and monitor providers, including default changesGV.SC-04; GV.SC-07SR-3 Supply Chain Controls and Processes; CA-7 Continuous MonitoringControl 15A.5.22
Codify secure defaults as your baselinePR.PS-01CM-2 Baseline Configuration; CM-6 Configuration SettingsControl 4 Secure Configuration of Enterprise Assets and SoftwareA.8.9 Configuration management
Security requirements in acquisitionGV.SC-05SA-4 Acquisition Process; SA-8 Security and Privacy Engineering PrinciplesControl 15A.5.23 Information security for use of cloud services

Decision checklist

  • Do we know which provider defaults apply to which of our accounts and resources, and from what date?
  • Is every control that a default satisfies still owned by a named person on our side?
  • Are the defaults we rely on enforced by our own policy, so they survive a provider change?
  • Does our contract give notice of changes to relied-on defaults and define incident cooperation?
  • If we use blueprints or compliance packages, is the version pinned and the fork reviewed?
  • If we share posture with insurers, do we understand the policy conditions about maintaining it?

Practitioner checklist

  • Build a per-provider register of secure defaults with scope and start dates.
  • Back-fill defaults to pre-existing accounts, organisations and subscriptions with your own policy.
  • Test new guardrails in audit or dry-run mode, then pilot, then enforce, with a rollback route.
  • Run a time-limited, owned exception process for any removed default.
  • Alert on organisation policy, Block Public Access, metadata option and policy assignment changes.
  • Add a responsibility schedule and default-change notice clause to critical provider contracts.
  • Review provider SOC 2 or ISO/IEC 27001 reports and implement the user entity controls they list.

Standards and references

Related NHI Mgmt Group resources: Shared Responsibility Models Guide · Shared Responsibility Matrix Template · Cloud Security Guide · Cloud Landing Zone Security Guide · Cyber Insurance Guide

PDF for subscribers

Take this guide with you

Every section, table and checklist in one PDF, with links to the sources and related guides. Subscribe to our weekly NHI & AI Identity Journal and we’ll email you a download link.

Free weekly newsletter

Subscribe to the NHI & AI Identity Journal

The latest on NHI and Agentic AI security – articles, research, breaches, news and events every week.

Bonus 33% off our NHI Course when you subscribe.

NHIMG Editorial Note
Researched and written by the NHI Mgmt Group editorial team using AI tools, under the editorial direction of Lalit Choda. Facts are drawn from the sources listed on this page and checked by our automated quality process. Last updated 14 November 2026. Spotted an error? Use our contact page.
NHI Mgmt Group, the independent authority on Non-Human Identity, IAM and Agentic AI security. nhimg.org

PDF Downloads for Newsletter Subscribers

Get the guide

Enter your details and we’ll email you a download link.

The full guide is free to read online. The PDF is a thank you for subscribing.

Fill in all fields marked * and tick the PDF box to continue.