Join our Newsletter — 33% off our NHI Course
Home› Glossary› Cyber Security› Maker-Checker Split
Cyber Security

Maker-Checker Split

← Back to Glossary
By NHI Mgmt Group Updated October 31, 2026 Domain: Cyber Security

A control pattern that separates code generation from code validation so the same system does not approve its own work. In agentic development, this reduces bias, improves auditability, and creates a defensible review trail for engineering, security, and compliance teams.

What Maker-Checker Split Means in Practice

Maker-checker split is a control architecture, not just a workflow choice. It separates creation from approval so one actor, system, or agent cannot both produce and validate the same change. That separation matters most where code, policy, configuration, or permissions can create downstream business or security impact if they are accepted uncritically.

The core value is independence. A maker optimises for producing the change, while the checker is expected to challenge it, test it, or verify it against policy. In agentic development, that boundary is especially important because the same automation that accelerates delivery can also amplify mistakes, hidden assumptions, or overconfident outputs if review is not genuinely independent.

Where Maker-Checker Split Adds Security Value

In security-sensitive engineering, maker-checker split reduces the chance that flawed logic, unsafe configuration, or a privileged action is self-approved. It creates a review trail that can show who proposed a change, who validated it, and whether the checker applied meaningful scrutiny rather than rubber-stamping the output.

The pattern is most useful when the generated work can affect access, data handling, release integrity, or production behaviour. That is why controls such as dual control and four-eyes approval for privileged and agent actions are a natural fit for destructive changes, while the broader AI Agents in Financial Services Security Guide shows how supervisory review, transaction limits, and audit evidence become central when AI agents act inside regulated workflows.

Used well, the pattern supports accountability as much as it supports security. It helps engineering, security, and compliance teams prove that high-impact work was reviewed by someone other than the originator, which is often the difference between a defensible control and a process that only appears controlled.

Maker-Checker Split in Agentic Development

In agentic development, the maker may be a developer, an automated pipeline, or an AI agent that drafts code, policies, tests, or infrastructure changes. The checker then acts as an independent control point, validating that the generated output is correct, appropriate, and authorised before it reaches a higher-trust environment.

This matters because agentic systems can produce plausible but wrong output at scale. A good split forces the organisation to decide which changes require human review, which can be auto-validated, and which should be blocked until a separate reviewer confirms intent, correctness, and policy alignment. The review step is not only about correctness, it is also about containing delegated authority.

That separation maps closely to preventing self-review in approval workflows, especially where the same toolchain can generate, test, and deploy code. It also aligns with the NIST view of access control and auditability, where independence and traceability are both part of sound control design.

What Makes a Defensible Review Trail

A defensible review trail records more than a timestamped approval. It should show the identity of the maker and checker, what changed, why it changed, what evidence was used to assess it, and whether any exceptions were granted. Without that context, approval becomes a formality rather than a control.

Auditability improves when the checker is given enough signal to detect drift, policy violations, or unsafe reuse of prior decisions. That is why maker-checker works best when paired with clear change descriptions, diff visibility, policy checks, and immutable logs that preserve the sequence of events leading to approval.

For teams working with AI-generated artefacts, the same discipline also helps distinguish legitimate automation from unreviewed machine output. A review trail only adds value when it can be trusted to reflect real challenge, not a mechanically repeated endorsement.

Operational Limits and Control Design Trade-offs

Maker-checker split is not a universal answer. It slows delivery, creates dependency on reviewer quality, and can fail if approvers lack context or authority. If the checker is too detached from the work, the control degrades into a box-tick; if the checker is too close to the maker, independence is lost.

The strongest designs reserve the pattern for material changes, high-privilege actions, and sensitive workflows where the cost of a bad decision is high. Lower-risk activity may be better handled with automated policy enforcement, peer review, or testing gates, while the most sensitive actions need a genuinely separate approver with the right expertise to challenge the change.

In that sense, maker-checker split is a governance pattern as much as a technical one. It works only when independence, authority, and evidence are all designed into the workflow, not added as afterthoughts.

Standards & Framework Alignment

This section maps relevant standards and security frameworks to the operational risks and controls described in this guidance.

OWASP Agentic AI Top 10 addresses the attack and risk surface, while NIST SP 800-53 Rev 5 and OWASP ASVS set the governance and control requirements practitioners need to meet.

FrameworkControl / ReferenceRelevance
NIST SP 800-53 Rev 5AC-3 — Access EnforcementSeparates who can approve or execute privileged changes.
AU-2 — Event LoggingMaker-checker needs a durable trail of who proposed and who approved the change.
CM-3 — Configuration Change ControlThe pattern is a change-control safeguard for code and configuration updates.
Recommendation — Enforce AC-3 so the reviewer and actor have distinct approval and execution authority. Record maker and checker actions under AU-2 to preserve review accountability. Apply CM-3 to require independent approval before implementing material changes.
OWASP ASVSV15 — Secure Coding and ArchitectureIndependent review strengthens secure design and prevents self-approved flaws.
Recommendation — Use V15 review gates to keep code authorship and validation separate.
OWASP Agentic AI Top 10ASI03 — Identity & Privilege AbuseMaker-checker reduces an agent's ability to self-authorize unsafe actions.
Recommendation — Require independent approval before an agent can exercise elevated authority.

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
Reviewed and updated by the NHIMG editorial team on October 31, 2026.
NHI Mgmt Group — the #1 independent authority on Non-Human Identity, IAM, and Agentic AI security. nhimg.org