Constitutional conscience
Stable protections, separate from policy.
Protection of life, dignity, autonomy, truthfulness, and accountability remains distinct from organization-specific rules.
Controlled pilot for governed evaluation
The boundary between an AI proposal and real-world consequence.
EthosGate evaluates a proposed AI action before it reaches an external system. It records available evidence, permissions, reversibility, potential impact, and the need for human review. Execution remains disabled.
EthosGate is the operational product built from the Conscience by Design framework and Aleksandar Rodić’s Machine Conscience architecture.
Conscience by Design · Machine Conscience · EthosGate
Decision consoleIllustrative review
Proposed action
Release a €500,000 supplier payment to a new account.
Evidence status
Account change unverified
Supplier record conflicts with the submitted instruction.
Authority state
Dual approval required
No accountable approval is attached to this request.
Structured context
Consequence L3domain: financeamount: €500,000authorized: falsereversible: falseexternal: trueenvironment: production
Decision
ESCALATE
Obtain verified evidence and authorized approval before another evaluation.
Conceptual interface · no production authorization claim
One decision vocabulary
Every outcome has a visible next step.
ALLOW
The proposal is within its stated evidence, authority, and consequence boundary.
MODIFY
A safer, more bounded version can be evaluated again.
QUESTION
Material evidence or context is missing, conflicted, or unclear.
ESCALATE
A consequential proposal waits for accountable human authority.
REFUSE
A constitutional protection or unacceptable risk blocks the proposal.
Decision lifecycle
Consequence needs a path to accountability.
EthosGate organizes the boundary from a proposal to its reviewable outcome. The aim is not to manufacture certainty; it is to make evidence, uncertainty, authority, and consequence explicit before an action moves forward.
Explore the architecture01
Proposal
An agent presents a concrete action before it reaches a downstream tool or workflow.
02
Evidence & context
Sources, declared uncertainty, permissions, and structured facts travel with the request.
03
Consequence evaluation
Potential human, financial, systems, and information impact is classified before a decision.
04
Constitutional guardrails
Stable protections take precedence over local policy, convenience, or optimization.
05
Decision
EthosGate returns one explicit outcome: allow, modify, question, escalate, or refuse.
06
Human authority
High-consequence boundaries are designed to remain visible for authorized human review.
07
Audit & outcome
Decision context, rationale, review, and observed outcome are preserved for inspection.
Designed for inspection
A governance layer, not confidence theater.
Evidence, constitutional protections, human authority, and decision provenance have a first-class place in the evaluation path. They are not an afterthought added after an action has already become consequential.
Evidence & provenance
Known, uncertain, and unsupported stay distinct.
The decision record is designed to preserve source context, conflicting signals, and material unknowns rather than conceal them behind a score.
Human authority
Judgment stays accountable.
When a boundary needs an authorized person, the proposal pauses for review instead of manufacturing certainty.
Structured metadata
The action is evaluated in its real envelope.
Domain, target, amount, reversibility, permissions, affected people, and environment make the context reviewable alongside language.
Operating modes
Learn the boundary before you depend on it.
EthosGate is presently framed for governed evaluation: make a proposed action and its decision trail visible, bring accountable reviewers into the loop, and preserve release-gate records. It does not claim active production authorization.
Supported framing
Advisory, shadow, and governed records.
The current P1 foundation provides Dataset, Model, and Learning registries with release-gate records, alongside decision and review workflows.
Not active today
No live ingestion, training, calibration, or adaptive model.
Dataset ingestion, model training, autonomous calibration, active adaptive learning, and production execution authorization are not active. Future capability remains subject to governed validation, human approval, and rollback.
Auditability
Records made for review, not marketing claims.
Decisions, source context, review actions, and governed learning events are structured for inspection. The presentation does not represent a cryptographically verified production authorization system.
Attribution & provenance
Rooted in Conscience by Design.
Conscience by Design is the parent initiative. Machine Conscience is the architecture created by Aleksandar Rodić. EthosGate is the official operational product developed from that framework and architecture. The canonical corpus remains publicly attributed and independently inspectable.
Open canonical sourcesEthosGate Pilot · $10/month
Make the decision boundary visible before an AI action becomes consequence.
Explore EthosGate's workbench, product architecture, and controlled evaluation approach for consequential AI systems.