Skip to content
DAMREV
Start a Structuring Review

Guide

Four Eyes & Six Eyes

At DAMREV, no one both does a piece of work and approves it. Four eyes and six eyes determine who may assess, approve or review each class of decision, and the same rule binds DAMREV's people, its clients' people and the agentic artificial intelligence (AI) working inside the platform. Every approval is recorded against its inputs.

The principle

What is the four-eyes principle?

It is the rule that a consequential decision needs two people: one who prepares or assesses it and a second, independent person who approves it. Neither can complete the decision alone, so a single mistake, a single lapse of judgement or a single bad actor cannot carry it through.

  • DefinitionFour eyes, definedA control under which no decision of a defined class takes effect until a second person, independent of the first, has approved it. Banks call the same idea dual control or maker and checker, and auditors call the wider discipline segregation of duties.
  • Condition oneIndependent of the firstThe second person must be independent of the first, which means they did not prepare the work and do not report to the person who did.
  • Condition twoThe same inputsThe second person must see the same inputs, not a summary chosen by the first. A second signature on a document the approver never examined is a formality, not a control.
  • Why it existsA second pair of eyesThe principle is old because the failure it prevents is old. A person checking their own work sees what they meant to do, not what they did. A person acting alone can be pressured, can be wrong without knowing it, or can act in bad faith, and nothing in a single-signature process will notice. A second person with the same information and a different role breaks each of those failures.
  • Across financeTwo signaturesIt is common across finance for exactly this reason. Payment releases, credit approvals, changes to a securities register and the release of client money are routinely placed under two signatures. The control is cheap relative to the losses it prevents, and it produces evidence as a by-product: the record shows two names, not one.
  • On DAMREVA rule of the platformDAMREV applies the principle as a rule of the platform rather than a habit of its staff. Roles and access are held in Stratum, the operating core, and the approval path a decision must follow is set by its class. The decision cannot be completed by the same role that prepared it.
  • The regulator's positionFormat does not change the lawThree SEC divisions issued a Statement on Tokenized Securities on 28 January 2026. It separates issuer-sponsored from third-party-sponsored tokenized securities, and states that the format in which a security is issued does not affect the application of the federal securities laws1.
  • On a public ledgerThe approval comes before the stepA Stellar ledger closes roughly every five seconds with finality at close2, so the moment for a second opinion is before the transaction, not during reconciliation.
  • Tokenized programmesStill two peopleTokenization changes where the record lives and how fast a step settles, not whether a decision needs a second person. The same network lets an issuer set authorisation flags on an asset natively2, so an issuer's permission can be enforced on the ledger itself rather than only on paper. Encoding does part of the work that approvals used to do by hand. Sentinel applies the transfer rules at each transfer, in contract logic on Soroban, which has run natively on Stellar mainnet since February 20243, and a transfer that fails them does not proceed. What the platform cannot encode is the judgement that decides what the rules should be. That judgement is where four eyes and six eyes apply.

In a programme

Where do the controls apply across a programme?

At every stage of the fixed sequence, wherever work from one stage is relied on by the next. The asset and the claim are assessed before a structure is drawn; the structure is approved before rules are encoded; issuance and every later servicing event follow the approval path set for their class.

  1. The asset and the claim

    Assessed, then approved.

  2. The structure

    Vehicle, ranking and control framework approved.

  3. The instrument

    Rights and rules checked before encoding.

  4. Issuance and the register

    Minted under the licensed entity's authority. Issuance is performed by the licensed entity for the programme, which is not DAMREV. The approval path records the decision and the authority under which the instrument was minted, so the register can always be traced back to it.

  5. Servicing for life

    Each lifecycle event on its approval path.

Each class of decision in that sequence carries its own requirement: four eyes, or six eyes where a reviewer is added. What stays constant is that the approval path is set by the class of the decision, not chosen by the person making it, and that nobody can approve work they prepared.

What this page does not publish

The table that assigns each class of decision to four eyes or six eyes, and the definition of the floor below which a reviewer is added, are not published on this page. Ask for the version that governs a programme through the contact page.

Six eyes

What does six eyes add to four eyes?

A third person, who reviews. Under six eyes an assessor drafts, a separate approver signs off, and a reviewer independent of both examines the decision before it takes effect. On DAMREV the reviewer is added on adverse or below-floor decisions, where the consequence of a mistake is greatest.

Follow one decision. Assessment, then Approval, then The programme record. Choose an approval path and follow one decision through it.
  • Follow one decision
  • Assessment: Establishes the facts and drafts the finding, with its inputs attached.
  • Approval: Decides whether the programme may act on the finding, and is named on the record.
  • The programme record: Each assessment, approval and review is attached with its inputs.
Follow one decision. Assessment, then Approval, then The programme record. Choose an approval path and follow one decision through it.

Choose an approval path and follow one decision through it.

  • The reviewerA different purposeThe third pair of eyes serves a different purpose from the second. The approver asks whether the assessment is right. The reviewer asks whether the process that produced the assessment and the approval was sound: whether the right inputs were used, whether the right people acted, and whether anything was missed that neither of them would see from inside the decision.
  • When it appliesWhere it earns its placeAdding a reviewer to every decision would slow a programme without making it safer. Most decisions are routine, and two independent people are enough. The reviewer earns their place where a decision goes against the programme or falls below a threshold the method sets, because those are the decisions in which pressure to reach a convenient answer is greatest.
  • Agentic AIThe same rule binds agentsThe same rule binds agentic AI. Agentic AI works within the same defined roles and approval paths as people. Praxis, the agentic engine, runs agents that use the other engines as tools, and every agent action is routed through Sentinel and Veritas, signed via Sentry, and composed through the roles Stratum holds. An agent inherits a role; it does not create one.
  • AI governanceAligned, not certifiedPraxis is aligned with the model of recognised AI governance frameworks, including ISO/IEC 42001, the AI Risk Management Framework of the US National Institute of Standards and Technology (NIST), and the European Union's AI Act. It is aligned with them, not certified against them.

The record

How is each assessment, approval and review evidenced?

Each one is attached to the programme record with its inputs. An auditor can see not only that a decision was approved, but by whom, on what information, and under which version of the Standard, so the operating instrument can always be traced back to the terms that authorised it.

  1. 01

    The class of decision, and the approval path it required.
  2. 02

    The assessor, and the assessment as drafted.
  3. 03

    The approver, and the approval as given.
  4. 04

    Where a reviewer was added, the reviewer and the review.
  5. 05

    Where an agent acted, the role it acted in and the signature that authorised the action.
  6. 06

    The inputs each person saw, and the version of the Standard in force.

On request

Which documents can I ask for about approvals?

Ask through the contact page and name the programme and your role. Each box below opens the contact page for that request.

Common questions

What else do people ask?

Short answers on four eyes and six eyes: how they differ, whether a senior person can assess and approve alone, whether the controls apply to DAMREV's own staff, whether an AI agent can approve, which decisions need six eyes, how an auditor confirms two people acted, and whether Praxis is certified under an AI standard.

What is the difference between four eyes and six eyes?

Four eyes needs two people: an assessor and a separate approver. Six eyes adds a third person who reviews, and on DAMREV that reviewer is added on adverse or below-floor decisions.

Can the same person assess and approve if they are senior enough?

No. The person who builds a thing never signs it off, whatever their seniority. The approval path is set by the class of decision, not by the person making it.

Do the controls apply to DAMREV's own staff?

Yes. The same approval paths bind DAMREV's people, its clients' people and agentic AI. DAMREV runs its own operations on the platform, with Stratum as tenant zero.

Can an AI agent approve a decision?

An agent acts only in the role it is given in Stratum, and it cannot approve work prepared in its own role. Its actions are routed through Sentinel and Veritas and signed via Sentry on the user's device.

Which decisions need six eyes?

Adverse or below-floor decisions add a reviewer. The table that assigns every class of decision to four or six eyes can be requested through the contact page for the version that governs a programme.

How would an auditor confirm that two people really acted?

From the programme record. Each assessment, approval and review is attached with its inputs and the names of the people who acted, under the version of the Standard in force at the time.

Is Praxis certified under an AI standard?

Praxis is aligned with the model of recognised AI governance frameworks, including ISO/IEC 42001, the NIST AI Risk Management Framework and the European Union's AI Act. It is not certified against them.

Where next

The pages below pick up where this one ends. Start with the one closest to where your asset or programme stands today.

References

  1. 1.US Securities and Exchange Commission, Statement on Tokenized Securities, 28 January 2026. https://www.sec.gov/newsroom/speeches-statements/corp-fin-statement-tokenized-securities-012826-statement-tokenized-securities
  2. 2.Chainstack, "What is Stellar". https://chainstack.com/what-is-stellar/ Also: Stellar developer documentation, "Control access to an asset", https://developers.stellar.org/docs/tokens/control-asset-access
  3. 3.Stellar Development Foundation, "Protocol 20 and smart contracts are live on Mainnet". https://stellar.org/blog/developers/protocol-20-and-smart-contracts-are-live-on-mainnet

Last updated:

Ready to build on programmable infrastructure?

Tell us what you want to tokenize. We will show you how DAMREV can run it.