Compliance

CIS Controls v8.1: The 18 Controls Prioritized

Prioritize the 18 CIS Controls by dependency, assign evidence owners, and turn IG1 from a checklist into an operating program.

Alex Gibson, Cofounder and Principal at Artemes AI
Alex Gibson
Cofounder, Principal
Aug 31, 2026 9 min read
Dependency stack organizing the 18 CIS Controls into foundation, protection, detection, response, and validation layers

CIS Controls do not fail because 18 is too many. They fail when an organization treats the list as 18 separate projects and never builds the shared evidence underneath them.

The right starting point is not Control 1 because it has the lowest number. Start with the dependencies that make later controls possible: accurate assets, approved software, secure configuration, owned accounts, and reliable recovery. A penetration test cannot compensate for an asset inventory nobody trusts.

Version 8.1 gives teams 18 control areas and 153 safeguards. The value is prioritization. The work is deciding what applies, who owns it, what evidence proves it, and which result forces action.

Infographic

The CIS Controls are a dependency stack

Inventory and configuration support protection. Protection supports detection. Detection supports response and validation.

CIS Controls implementation dependency stackFour stacked layers group the 18 CIS Controls into know and govern, protect and recover, detect and respond, and validate and improve. Arrows show that higher layers depend on evidence from the layers below.VALIDATE AND IMPROVE13 Network monitoring | 16 App security | 18 Penetration testingDETECT AND RESPOND8 Audit logs | 10 Malware defense | 17 Incident responsePROTECT AND RECOVER3, 6, 7, 9, 11, 12, 14, 15KNOW AND GOVERN1 Assets | 2 Software | 4 Configuration | 5 AccountsStart at the dependency, not the control with the loudest finding

What are the CIS Controls?

The CIS Critical Security Controls are a prioritized set of defensive practices for common attacks. Each control states an outcome, and its safeguards break that outcome into measurable actions. They cover assets, software, data, configuration, identity, vulnerabilities, logging, recovery, networks, people, suppliers, application security, incident response, and testing.

CIS publishes an official CIS Controls v8.1 list with 18 controls. Version 8.1 keeps the structure from version 8 while aligning security function mappings with NIST Cybersecurity Framework 2.0, refining asset classes, and clarifying safeguard language.

Controls are not benchmarks. A control says to establish and maintain secure configuration. A benchmark gives product values for Windows, Linux, a database, a cloud service, or another technology. Our CIS Benchmarks guide explains profiles, versions, and exceptions. The system hardening guide shows how to turn those values into tested, observed state.

How do CIS Implementation Groups change the work?

Implementation Groups keep the framework from becoming an all at once program. IG1 is the minimum set for every enterprise. IG2 adds safeguards for greater operational complexity and risk. IG3 adds safeguards for organizations protecting sensitive systems against sophisticated attackers.

CIS's April 2022 essential cyber hygiene guide counts 56 safeguards in IG1, 74 additional safeguards in IG2, and 23 more in IG3, for 153 total. IG1 is about 37 percent of the full set. That is small enough to start and large enough to require real ownership.

Do not choose IG3 because the organization wants to sound mature. Choose it when the data, service consequence, threat, and available expertise justify the added control burden. Every safeguard creates evidence and change work. An unstaffed safeguard is a claim, not protection.

How should the 18 CIS Controls be prioritized?

CIS numbers provide a useful order, but implementation should follow dependencies. Use four operating phases.

Phase 1: know what exists and who can change it

  • Control 1, Inventory and Control of Enterprise Assets. Establish identity, owner, role, and connection state for devices and cloud resources.
  • Control 2, Inventory and Control of Software Assets. Record installed and running software, support status, version, publisher, and authorization.
  • Control 4, Secure Configuration of Enterprise Assets and Software. Assign current baselines, deploy them, and detect unexplained drift.
  • Control 5, Account Management. Find human, administrator, and service accounts, then tie every account to a current owner and purpose.
  • Control 6, Access Control Management. Grant only required access, review privilege, and remove it when the role or relationship ends.

These five controls produce the join keys for the rest of the program. If a vulnerability finding cannot map to an asset and owner, Control 7 cannot route it. If an audit event cannot map to an account, Control 8 cannot explain it.

Phase 2: reduce common paths and preserve recovery

  • Control 3, Data Protection. Identify sensitive data, control handling, encrypt it, limit access, and remove it when retention ends.
  • Control 7, Continuous Vulnerability Management. Discover flaws, match affected state, route action, and verify remediation.
  • Control 9, Email and Web Browser Protections. Reduce malicious content, unsafe extensions, weak domain controls, and risky web behavior.
  • Control 10, Malware Defenses. Prevent or control malicious execution and confirm protection remains active.
  • Control 11, Data Recovery. Protect backups and prove important services can return to a trusted state.
  • Control 12, Network Infrastructure Management. Inventory network devices, secure management, control services, and maintain current configurations.

Recovery belongs in this phase, not at the end. A team that changes security settings without tested recovery will avoid necessary changes after the first outage. Control 11 gives the program room to act.

Phase 3: create detection, human readiness, and external accountability

  • Control 8, Audit Log Management. Collect the events needed to investigate identity, system, and control change, then prove ingestion and retention.
  • Control 13, Network Monitoring and Defense. Observe traffic and defensive signals across the network, with tuned thresholds and owned response.
  • Control 14, Security Awareness and Skills Training. Train people for the decisions their roles actually require, not one generic annual presentation.
  • Control 15, Service Provider Management. Inventory material providers, define security obligations, review evidence, and plan exit.
  • Control 17, Incident Response Management. Assign authority, rehearse decisions, preserve communications, and learn from exercises and real events.

Phase 4: validate the program where failure costs the most

  • Control 16, Application Software Security. Build secure development, dependency management, testing, and defect correction into software delivery.
  • Control 18, Penetration Testing. Test whether controls stop realistic attack paths and feed the result back into owners and engineering.

This order is judgment, not a replacement for the official Implementation Groups. It exposes dependencies. An organization can move a control earlier when its business demands it, but it should state what evidence the control depends on and how that evidence will be supplied.

What changed for CIS Controls in the last year?

The framework itself stayed stable, while the mapping layer improved. On August 26, 2026, CIS published an updated v8.1 mapping to NIST SP 800-53 Revision 5. This matters to regulated teams that implement one technical practice but must explain it through several frameworks.

A mapping is a translation aid, not inherited compliance. Implementing a CIS safeguard does not automatically prove every mapped NIST requirement. The scope, assessment method, parameters, and evidence can differ. Use the mapping to reuse evidence where claims align, then document gaps.

Who should own the CIS Controls?

The CISO can own the framework. The CISO cannot own every control result. Assign one accountable business or technology owner, one operating owner, and one evidence source per safeguard.

Control areaAccountable ownerEvidenceFailure trigger
Assets and softwareIT operationsObserved inventory with owner and ageUnknown asset or stale evidence
Identity and accessIdentity leaderEffective membership and access reviewOwnerless or excessive privilege
Vulnerabilities and configurationPlatform ownerAffected state, action, and verified closureExposed finding beyond target
Recovery and responseService ownerExercise result and recovery proofFailed test or expired plan

Keep security as the challenger and integrator. It should define evidence quality, test assumptions, track cross control gaps, and escalate decisions. When security becomes the owner of every stale asset and backup failure, operating teams learn that the framework is somebody else's problem.

How do you measure CIS Controls without checkbox reporting?

CIS publishes a Controls Assessment Specification with inputs, operations, measures, metrics, and procedure review for safeguards. That is the right model. State what data enters the measure, what operation transforms it, and which threshold creates action.

Coverage alone is weak. If 980 of 1,000 assets report to inventory, coverage is 98 percent. That sounds good until the missing 20 are internet facing appliances. Add evidence age, business role, exposure, and owner. Twenty unknown edge systems deserve more attention than 20 missing lab laptops.

Simple capacity math also changes planning. Fifty six IG1 safeguards with one accountable owner review each quarter creates 224 reviews a year. At 30 minutes per review, that is 112 hours before remediation. Automate evidence collection and spend those hours on failed controls, exceptions, and decisions.

Deep endpoint context with AI driven analysis can help join configuration, software, vulnerability, identity, and exposure evidence. It should not invent control status. Every conclusion still needs a source, observation time, and clear path to verification.

What should a safeguard evidence packet contain?

Use a small, repeatable record. Name the safeguard, scope, accountable owner, operating owner, approved method, source systems, observation time, measure, threshold, result, exceptions, and next review. Link the record to raw evidence rather than pasting screenshots into a document.

Distinguish design evidence from operating evidence. A policy and approved procedure show that the organization designed a control. Inventory results, configuration values, backup tests, access reviews, and exercise records show whether it operated. Many audits go badly because a strong policy is offered as proof of current behavior.

Preserve the denominator. Reporting 950 protected assets is meaningless unless the organization can defend why 1,000 assets were in scope. Keep the inventory query, filters, exclusions, and observation time. When the denominator changes, explain whether the fleet changed or the collection failed.

Give every failed measure a disposition: correction, approved exception, accepted risk, false match, or scope change. Require evidence for closure. A ticket marked done should point to a later observation that passed the same test. This keeps framework reporting tied to real state.

What does a practical 90 day CIS Controls rollout look like?

Days 1 through 30: establish truth

Select IG1, name executive and operating owners, and inventory current data sources. Measure asset coverage, software coverage, administrator membership, baseline assignment, backup scope, and log ingestion. Do not buy a tool to hide missing ownership.

Days 31 through 60: close dangerous gaps

Prioritize unknown internet assets, unsupported software, ownerless privilege, exposed services, failed backups, and missing security logs. Use known exploitation evidence and business consequence to order vulnerability work.

Days 61 through 90: make evidence repeatable

Automate collection, publish evidence age, define failure thresholds, and connect each failed measure to an owner and correction target. Run one recovery exercise and one incident exercise. Record what failed and fund the fix.

At day 90, leaders should be able to see which IG1 safeguards are implemented, which are failing, which lack evidence, and which have approved exceptions. A percentage without those four states is decoration.

Frequently asked questions

Are the CIS Controls free?

The Controls and supporting public resources are available from CIS. Some implementation tools, assessment content, and membership services have separate access or licensing terms.

What is the difference between CIS Controls and CIS Benchmarks?

Controls define program outcomes and safeguards. Benchmarks define secure configuration recommendations for specific products. Control 4 may lead you to a Windows or Linux Benchmark, but the documents serve different jobs.

Should a small organization implement all 153 safeguards?

Usually not at first. CIS says every enterprise should start with IG1. Add IG2 or IG3 safeguards when business consequence, data sensitivity, operating complexity, or threat justifies them.

Do CIS Controls prove compliance with NIST or CMMC?

No. Mappings help reuse aligned evidence, but each framework has its own scope, parameters, assessment method, and documentation. Validate the exact claim before declaring coverage.

Executive takeaway

Start with IG1 and demand evidence, not an implementation percentage. For every safeguard, name the accountable owner, current input, failure threshold, correction path, and evidence age. Then fix the five dependencies that support everything else: assets, software, configuration, accounts, and recovery. The CIS Controls become a program when failed evidence creates owned action.

Artemes AI

Put more evidence behind vulnerability decisions

Artemes AI combines endpoint telemetry, sourced vulnerability intelligence, and analysis with practitioner review so teams can examine the evidence, missing context, and recommended next step together. We are accepting early access requests now.

Alex Gibson, Cofounder and Principal at Artemes AI

Alex Gibson

Cofounder, Principal

Alex writes about configuration drift, operational security evidence, endpoint telemetry, triage supported by AI, and the practical work of turning signals into better remediation decisions.

Blue Team
Threat Modeling
Security Automation
Found this useful? Share it.

Get articles like this in your inbox.

Security research and occasional Artemes AI product updates.