Small businesses move to the cloud expecting the provider to handle security. The reality is a split that changes with every service tier. I have seen teams assume encryption, backups, and log alerts are automatic, only to find gaps during an audit or after an incident. This guide shows where the provider's duty ends and yours begins.
You will learn the provider's core responsibilities, the controls you must manage, the AI-specific boundaries, and the mistakes that leave organizations exposed. At Junious Digital Labs, we help clients map these matrices, compare providers, and stay the single point of contact for their technology stack. A free 30-minute call with Joshua can clarify the first steps.
How the Cloud Shared Responsibility Model Works
The cloud shared responsibility model explained simply: security and compliance are divided between the provider and the customer (AWS). This split also extends to IT controls (AWS). The responsibility to operate the IT environment is shared, so the management, operation, and verification of IT controls is shared as well (AWS).
The provider secures the physical facilities, power, cooling, and the hypervisor that runs virtual machines. You secure the guest operating system, applications, data, identity policies, and network traffic rules you create. With a managed database, the provider patches the engine; you still control the data, access lists, and encryption keys. With serverless functions, the provider manages the runtime; you own the code, permissions, and event triggers.
Common mistake: assuming the provider handles backups, encryption at rest, or log retention by default. They often do not. Common mistake: leaving the default security group open to the internet. Common mistake: never testing a restore.
What Your Cloud Provider Actually Secures
Providers run the physical data centers, power, cooling, and network backbone. They harden the hypervisor layer and patch the host operating systems. Managed services such as managed databases, identity platforms, and serverless functions include operating system updates and service-level encryption at rest. The provider also maintains the global fiber paths and edge locations that carry your traffic.
You still own the guest operating system, applications, data, and identity configuration. Misconfigured storage buckets, weak access policies, and unpatched application code remain your responsibility. The provider will not fix a leaked credential or an open security group you created.
Check the provider responsibility matrix for each service tier you consume. Confirm which controls the provider documents for audit purposes. Ask your team to list every cloud resource, assign an internal owner, and document the configuration settings that owner manages.
Security Tasks You Still Own Completely
You own the data you place in the cloud. You decide who can reach it and what they can do with it. The provider does not manage your accounts, permission sets, or encryption keys. You must use a managed key service and set the policies yourself. You must define and document user roles and responsibilities with regard to external system services, including shared responsibilities with external service providers (NIST).
Turn on multi-factor authentication for every human and service account. Review group memberships each quarter. Remove access the same day a person leaves or changes roles. Encrypt data at rest and in transit using keys you control. Validate that storage buckets, databases, and queues are not open to the internet.
Harden the operating system images you build. Apply patches to your applications and guest operating systems on a schedule you set. Log all administrative activity and ship those logs to a separate account you monitor. Test your backup restores at least twice a year. Document the configuration settings you manage so an auditor can trace each control to an owner.
Shared Responsibility Areas for AI Workloads
Generative AI workloads move the boundary between provider and customer in ways that standard cloud models do not cover. Microsoft publishes a dedicated AI shared responsibility model for these workloads (Microsoft). The provider still secures the underlying infrastructure and the foundation model itself. You are accountable for your prompts, the data you submit, and the outputs you store or act on. Access controls, content filtering settings, and logging of model interactions are your decisions.
Autonomous agents that call APIs, query databases, or trigger workflows add another layer. Microsoft describes a separate AI agent shared responsibility model for that scenario (Microsoft). The provider manages the agent runtime and the tool connectors it offers. You own the permissions for agents, the scope of tools, and the audit trail that records all agent actions. Treat each agent like a privileged service account: limit its scope, monitor its activity, and review its decisions weekly.
Action step: list every AI service and agent in use. Map each control to the provider or to your team.
Common Mistakes That Leave Gaps Open
Many small teams assume the provider handles encryption by default. In practice the provider encrypts the disk but you must enable encryption for each bucket and database. Another common gap is identity. Businesses create one admin account and share the credentials across the staff. That single account becomes a blast radius if it is compromised. Logging is a third blind spot. Providers write the system logs but they do not alert on failed logins or data exports. You must configure the alert rules.
Backup is often confused with replication. Replication keeps a live copy in another zone; it does not protect you from ransomware that encrypts both copies. You need an immutable backup stored in a separate account. Patch management for virtual machines is also overlooked. The provider patches the hypervisor; you patch the operating system and applications inside the instance. Finally, network controls are left wide open. Security groups default to allow all outbound traffic. Restrict egress to only the destinations your workloads require.
Joshua Junious: "The most common gap I see is identity. Teams share one admin login, never rotate the key, and have no alert when that account does something unusual. Fix that first."
Weekly Checklist: Verify Compliance and Close Gaps
Start by listing every cloud service your organization consumes. Map each service to the provider documentation that defines the security boundary. Check identity and access management first. Review role assignments quarterly and remove any that are no longer needed. Validate encryption settings for data at rest and in transit. Ensure keys are managed in a dedicated key management service and that rotation policies are configured.
Run a network configuration review. Verify that security groups and network access control lists restrict inbound and outbound traffic to known endpoints. Disable any unused ports or protocols. Request the provider's latest compliance reports such as SOC 2 Type 2 or ISO 27001. Compare the control objectives in those reports against your own regulatory requirements. Document the findings in a simple tracker. Assign an owner and a due date for every open item. Schedule the next review before the current one expires.
This week: download the provider responsibility matrix for your top three services, audit every admin account for MFA and ownership, and test one backup restore from an immutable copy. For a related guide, see Move File Server to Cloud Checklist for Small Business. If you want help with this work, see our cloud and infrastructure services and data center and colocation services pages.
Book a free 30-minute call with Joshua to talk it through.




