Contact DA
ETOS Platform · v3.5 · governance-native

ETOS is no longer just a decision model —
it is a self-regulating infrastructure

For high-risk environments where learning, accountability, and governance must remain aligned.

ETOS is designed to surface tensions, keep decisions explainable and auditable, and allow the system to learn over time without stepping outside its constitutional limits.

The key shift

ETOS is now framed as a governance-native decision infrastructure in which system evolution is continuously shaped, constrained, and corrected under governance. It does not merely learn from the world — it learns from its own governance history.

Core architecture

Four layers, one closed governance loop

ETOS is built as a recursive architecture in which proposals, evaluation, historical memory, and active regulation are directly connected.

01 · ETOS-Self

Proposal Engine

Generates proposals from context, conflict analysis, and governance state. Output includes proposal, conflict profile, decision trace, and governance adjustments.

02 · CDP

Constitutional Decision Protocol

Evaluates proposals through technical, risk, and conceptual layers. Output includes integrity score, decision mode, governance flags, and ethical debt.

03 · Drift Memory

Reflection Center

Builds the system’s historical understanding of directional drift, velocity, acceleration, sequence patterns, and the lifecycle of ethical debt.

04 · Governance Layer

Active regulator

Holds the platform together through the Drift Pressure Index, oscillation detection, debt lifecycle management, and governance modes such as optimize, monitor, stabilize, and pause.

v3.0 → v3.5

How ETOS has evolved

v3.0

Drift-aware governance

CDP becomes historically aware. Drift Pressure Index and the ethical debt lifecycle are introduced. The system begins to understand its own history.

v3.1

Recursive feedback loop

ETOS-Self learns from CDP rejections. Penalties affect future proposals, and persistent feedback memory is introduced.

v3.2

Observability

Dashboard, feedback log, and drift visualisation make the system readable and inspectable.

v3.3

Proactive governance

Governance context is pulled into ETOS-Self before the decision is made. The system starts reasoning in context, not only through after-the-fact control.

v3.4

Dynamic behaviour

Velocity and acceleration awareness, together with a Governance Summary Layer, make the system aware of its own movement.

v3.5

Active steering

Governance becomes steering rather than merely observational. Recommended mode shapes future proposals, human review can be enforced, and the system can be actively stabilised.

Key concepts
T

Tensions

Decisions are analysed as tensions between values: efficiency vs precaution, throughput vs sensitivity, autonomy vs oversight, and other structural conflicts.

D

Drift

The system’s movement over time: direction, velocity, acceleration, and stability. Not only what happens now, but how the platform is changing.

E

Ethical Debt

The durable consequences of decisions. Debt accumulates per tension family, affects future decisions, and requires active resolution.

G

Governance Modes

Optimize, monitor, stabilize, and pause. Governance is not documentation afterward — it is an operational mode within the system.

Decision flow

A closed governance loop

Input → ETOS-Self → Proposal

CDP (Technical → Risk → Conceptual)

Decision + Debt + Flags

Drift Memory + Reflection

Feedback → ETOS-Self

Future system behaviour changes
historical accountabilityexplainabilityauditabilityactive stabilisation
EU AI Act · compliance

ETOS is not merely EU AI Act-compliant —
it is built as if compliance were a design requirement from day one

The EU AI Act requires explainability, human oversight, risk management, and auditability for high-risk AI systems. ETOS does not merely satisfy these requirements — its governance-native architecture is structurally identical to them.

Art. 13 · Transparency

Explainability is structural

ETOS generates a decision trace for every decision: which tensions were active, what governance state the system was in, and what influenced the outcome. Not as a log after the fact — as output within the process.

Art. 14 · Human Oversight

Human control is enforceable

Human review enforcement is an operational mechanism in ETOS v3.5 — not a recommendation. The governance layer can enforce human review when ethical debt is high, oscillation is detected, or critical tension is present.

Art. 9 · Risk Management

Risk is continuous, not point-in-time

The Drift Pressure Index and ethical debt lifecycle are ongoing risk measures that accumulate over time. ETOS satisfies the continuous risk management requirement with a mechanism designed for it — not retrofitted.

Art. 12 · Record-keeping

Historical accountability as a core layer

Drift Memory and the Reflection Center are built to preserve and analyse the system's historical decision behaviour. Auditability is not an add-on — it is one of the four architectural layers.

Art. 15 · Accuracy & Robustness

Active stabilisation under pressure

Governance modes (optimize, monitor, stabilize, pause) are operational responses to system state. The system can be actively stabilised — it does not depend on external intervention to avoid robustness failures.

Overall assessment

Compliance as consequence, not compromise

ETOS was not adapted to comply with the EU AI Act. It was designed from the same principles the Act attempts to codify. That is the difference between a system that can document compliance — and one that embodies it.

Art. 9 Risk Management Art. 12 Record-keeping Art. 13 Transparency Art. 14 Human Oversight Art. 15 Accuracy & Robustness
What makes ETOS different

The system regulates its own evolution through governance

ETOS treats decisions as tensions rather than answers. It learns under governance rather than around it. And it can explain what it does, why it does it, and how it changes over time.