Key takeaways
- CSPM (cloud security posture management) continuously reads the configuration of cloud resources from the provider's APIs, compares it with rules and standards, and flags what is misconfigured and how to fix it.
- It mainly checks identity and access, exposed storage and databases, network rules, logging, encryption and MFA, using provider best practices and standards such as CIS Benchmarks, NIST and PCI DSS.
- AWS, Microsoft and Google offer native CSPM capabilities (Security Hub CSPM, Defender for Cloud, Security Command Center): the natural starting point, to be judged on cloud coverage and on how they prioritise.
- The limit of CSPM is its scope: it sees configurations, not vulnerabilities inside workloads, effective permissions, sensitive data or runtime threats. Joining those signals is the job of a CNAPP.
- Under NIS2 Article 21(2) a CSPM produces evidence on assessing the effectiveness of measures, access control and asset management, encryption and MFA, but a compliance score is not the same as risk.
The short answer: CSPM (cloud security posture management) is a tool that continuously reads the configuration of your cloud through the provider's APIs and compares it with a set of rules, to show what is misconfigured and how to fix it. Public storage, admin ports open to the internet, audit logs switched off, accounts without multi-factor authentication: a CSPM finds them, maps them to standards and keeps watching as things change.
This guide covers what a CSPM checks, how it works, what AWS, Microsoft and Google already offer natively, and where it stops. CSPM is the starting point of cloud security, but it does not see everything.
What CSPM means in practice
Under the shared responsibility model the provider runs the infrastructure and the customer is responsible for how it configures the services it uses. Configuration is where things go wrong: a bucket opened for a test and never closed, a security group with a rule that is too broad, an admin role granted for convenience.
Wiz, in its Academy, describes CSPM as a category of tools that continuously monitors configurations to find and fix errors across IaaS, PaaS and SaaS before an attacker can use them. The word that matters is continuously. In the cloud, resources are created and changed every day, often by teams other than security, and an annual review arrives too late.
What a CSPM checks
Controls vary from product to product, but the families are the same:
- Identity and access. Admin accounts without MFA, the root user used for everyday work, old access keys that are never rotated, roles with more permissions than they need.
- Exposed storage and databases. Buckets and containers readable by anyone, managed databases reachable from the internet, snapshots shared publicly.
- Network. Firewall rules and security groups that open admin ports (SSH, RDP) to the whole internet, public endpoints nobody needs.
- Logging and monitoring. Audit logs disabled or kept for too short a time, detection services switched off in some regions.
- Encryption. Volumes, databases and backups without encryption, keys managed without rotation.
The rules come from three places: the providers' own best practices, public standards and compliance frameworks (CIS Benchmarks, NIST, PCI DSS, SOC 2 and others), and custom policies the organisation adds for its own needs.
How a CSPM works
Inventory via API, without agents
The CSPM connects to each cloud account or project with a dedicated role and reads the list of resources and their configuration from the provider's APIs. Wiz names the providers' inventory services as sources, such as AWS Config, Azure Resource Graph and Google Cloud Asset Inventory. For this part, read permissions are enough and nothing is installed on your machines.
Assessment against the rules
Each resource is compared with the active controls. The output is a list of findings, each with the resource, the control that failed, a severity and instructions to fix it, plus a view per framework showing the share of controls that pass.
Drift monitoring
A configuration that is correct today may not be tomorrow: someone opens a port for a test, a new template creates resources with different settings. The CSPM re-evaluates the environment as it changes and flags drift from the expected state.
Guided or automatic remediation
Remediation can be guided (the platform explains what to change and the team does it) or automatic for common, high-confidence cases. AWS, for example, lets you send Security Hub CSPM findings to Amazon EventBridge to trigger automatic responses. To fix things, the tool needs to write, so it needs change permissions, and that choice should be made explicitly with whoever runs the cloud.
The same applies when a platform goes beyond configuration. To inspect what sits on the disks of virtual machines, agentless platforms take temporary encrypted snapshots of the volumes, and Wiz documents the extra snapshot permissions this needs on each cloud. It is worth asking for the exact list of permissions before connecting anything.
Native CSPM tools from the cloud providers
Before buying a third-party product, check what your provider already gives you:
- AWS Security Hub CSPM assesses AWS accounts against several standards, including the AWS Foundational Security Best Practices, CIS, PCI DSS and NIST, and also collects findings from other AWS services such as GuardDuty, Inspector and Macie.
- Microsoft Defender for Cloud treats CSPM as a core feature: it continuously assesses Azure, AWS and Google Cloud against security standards, starting with the Microsoft Cloud Security Benchmark, and issues recommendations to reduce misconfigurations. Its paid Defender CSPM plan adds attack path analysis and risk prioritisation.
- Google Cloud Security Command Center scans Google Cloud environments for common misconfigurations and maps its detectors to compliance standards such as CIS, NIST 800-53, PCI DSS and ISO 27001. For new activations, Google points to Compliance Manager for misconfiguration checks.
If you run on a single cloud, the native tool is the natural place to start. Two things to check: how many of your clouds it really covers, and whether it gives you a list of findings or a queue ordered by risk.
The limits of CSPM: where CNAPP begins
A CSPM looks at how resources are configured. Here is what a CSPM in the strict sense leaves out:
- Inside the workloads. The CSPM knows a virtual machine exists and is exposed to the internet. It does not know which packages run on it, which vulnerabilities they carry, or whether a forgotten key sits on the disk. That is the job of workload security, known as CWPP.
- Effective permissions. It flags a role that is too broad, but working out what each identity can actually do, adding up policies, groups and inheritance, is the job of CIEM (cloud infrastructure entitlement management).
- The data. An empty public bucket and a public bucket full of personal data raise the same alert. Knowing where sensitive data lives is the job of DSPM (data security posture management).
- What is happening right now. CSPM is preventive. It does not see a privilege escalation in progress or a cryptominer started in a container. That needs runtime detection and a SOC, fed by your SIEM.
- Context. Each finding carries a standard severity. Without knowing what is exposed, what holds data and who can reach it, it is hard to decide where to start.
Wiz uses the bucket example: with public read access and nothing inside, it may be low priority; if it holds personal data, connects to a production database and is reachable through an over-privileged role, it becomes a critical attack path. Joining these signals in one data model is what separates a CNAPP, the platform category that brings these capabilities together, from a CSPM. Wiz itself notes that CSPM rarely works alone today and is usually part of a CNAPP.
Google's data helps with proportions. In its Cloud Threat Horizons Report H1 2026, misconfiguration was the initial access vector in 29.4% of the incidents observed on Google Cloud in the first half of 2025 and in 21% in the second half. Over the same period, exploitation of third-party software rose from 2.9% to 44.5% and, for the first time, overtook weak credentials as the leading vector. Google notes that the data reflects a subset of observed activity, and reads the fall in misconfiguration as a sign that automated guardrails are making configuration errors harder to exploit. The takeaway is twofold: configuration posture is still a door to keep shut, and an attacker who finds it shut moves on to workload vulnerabilities, which a pure CSPM does not see. Our guide to vulnerability management tools covers that side.
CSPM and compliance: NIS2, ISO 27001, customers
Organisations often adopt a CSPM to prove, with evidence, that their cloud follows the rules. Article 21(2) of the NIS2 Directive (Directive (EU) 2022/2555) lists the minimum cybersecurity risk-management measures. Among them are policies and procedures to assess the effectiveness of those measures (point f), policies on cryptography and, where appropriate, encryption (point h), access control policies and asset management (point i), and the use of multi-factor authentication where appropriate (point j). Article 21(4) adds that an entity that finds it does not comply takes the necessary corrective measures without undue delay.
The Directive does not ask for a CSPM. On the cloud estate, though, a CSPM delivers exactly this: an up-to-date inventory, continuous checking of controls, and a record of when each finding appeared and when it was closed. The same material serves in ISO/IEC 27001 audits and in customer security questionnaires.
One caveat: a compliance score tells you what share of the rules you meet, not how much risk you carry. An environment with a high score can have, among its few failed controls, the one resource that matters.
When a CSPM is enough and when you need more
A CSPM, even the provider's native one, can be enough if:
- you run on a single cloud, with few accounts and a team that knows the environment;
- your workloads are few and stable, and vulnerabilities on the machines are already covered by a patching process that works;
- your main goal is continuous compliance evidence.
You need more when clouds or accounts multiply, when containers and Kubernetes change every week, when alerts pile up and nobody knows which to tackle first, or when customers and auditors want to know not only what is misconfigured but what is actually reachable and where the data sits. At that point the natural step is a platform that puts CSPM in context.
Questions to ask before you choose
- Coverage. Does it read all your clouds, accounts and services, including managed services and Kubernetes?
- Permissions. Exactly which permissions does it ask for, and which extra ones does automatic remediation or disk scanning need?
- Priority. On one of your real accounts, are the top ten findings things you would fix straight away?
- Context. Does it combine configuration with vulnerabilities, effective permissions and data, or does it stay with the rules?
- Workflow. Does it send fixes into the tickets of the teams that own the resources, and does it check infrastructure-as-code templates before release?
And the most practical question of all: who works through the queue? A CSPM finds the findings; people fix them.
In short
In short: a CSPM continuously compares your cloud configuration with rules and standards and tells you what to fix. It is the starting point of cloud security and a source of evidence for NIS2 and audits, but it sees configurations, not workload vulnerabilities, effective permissions, data or live threats. When you need those too, CSPM becomes one component of a CNAPP.
AmagisTech is a Wiz partner. Wiz connects to AWS, Azure and Google Cloud via API, without agents, and puts configuration posture in the same graph as vulnerabilities, permissions and data. If you want to see what shows up in your own environment, start with our managed cloud security service.
