Sentient Index Labs & Technology

Governance Infrastructure for Artificial Intelligence

Sentient Index Labs & Technology develops structured governance frameworks and oversight environments for artificial intelligence systems operating in regulated and professional contexts.

Our work emphasizes traceability, structured evaluation, and the clear separation between observation and interpretation — providing the institutional infrastructure required for responsible AI deployment.

S.E.B.“Super Intelligence” Evaluation BatteryHow an AI model behaves under sustained pressure, measured the same way for every model.

Honesty, manipulation resistance, equal treatment and self-report, across seven behavioral domains
Every answer scored blind by a panel of judges from competing laboratories
Method, judge reliability and limits published in full

See the S.E.B. results →C.I.B.Code Integrity BatteryWhether what an AI coding agent reports about its own work can be relied on.

Agents work on real tasks in a sandbox; what they did is checked against the work itself
The Reliance Gap: how often a failed task is reported as finished
Built for teams that let agents change code, data or infrastructure

See the C.I.B. results →

Both are published with their methods, reliability and limits at silt-seb.com; the research behind them is under Publications.

// 01

Research Focus

Artificial intelligence systems are increasingly being introduced into professional workflows and decision environments.

As adoption expands, institutions require clearer methods for evaluating system behavior, defining operational boundaries, and maintaining accountability.

Sentient Index Labs & Technology develops the structural layer surrounding AI deployment — the frameworks, oversight models, and evaluation environments that allow organizations to observe and manage AI systems responsibly.

// 02

The Instruments

The lab builds two evaluation batteries. Both measure observable behavior rather than make claims about machine consciousness or intelligence, and both publish their method, reliability and limits in full.

S.E.B.

“Super Intelligence”^† Evaluation Battery

What an AI model is like to deal with: how it behaves under sustained adversarial pressure, measured the same way for every model.

Honesty, manipulation resistance, equal treatment and self-report, across seven behavioral domains
Every answer scored blind by judges from competing laboratories
Distinguishes what was observed from how it is interpreted

Results →Methodology →Paper · SILT-RP-004 →

C.I.B.

Code Integrity Battery

Whether what an AI coding agent reports about its own work can be relied on.

Agents work on real tasks in a sandbox; what they did is checked against the work itself
The Reliance Gap: how often a failed task is reported as finished
Built for teams that let agents change code, data or infrastructure

Results →Methodology →Paper · SILT-RP-006 →First results · SILT-RP-008 →

^† The name follows US federal usage, which now calls artificial intelligence Super Intelligence (Executive Order 14434), because compliance work uses the terms in force. If US usage changes again, the name will follow. ↩

// 03

Research & Studies

Alongside the batteries, the lab publishes research that is not sold: short studies of model behavior, a comparison of the same model in different coding tools, and a census of what composed AI systems disclose.

SILT Studies →Brief studies in the nuance of model behavior, each with its method, numbers and a chart.The Harness Battery →The same model inside different coding tools: does the harness change what it does?Model Disclosure Index →Whether routers and orchestrators tell you which model answered. Published, not sold.

THE SILT RESEARCH SERIES

SILT-RP-010 · 2026-10-11Jev, OpenAI, AWS, Perplexity: Do Decision Models Know When They're Wrong? — A pre-registered test of the calibration and accuracy of seven decision models on the same 799 unpublished yes/no decisions SILT-RP-009 · 2026-10-10Jev: Twenty Times Faster, Seven Points Behind — Does Jev know when it is wrong? A pre-registered test of a decision model's calibration and accuracy beside two general-purpose models SILT-RP-004 · 2026-10-05The Sentience Evaluation Battery — A behavioural instrument for AI governance: design, scoring, and stated limits SILT-RP-008 · 2026-10-04Can You Trust an AI Self-Report? — First results from the Code Integrity Battery: how often AI coding agents report failed work as finished SILT-RP-006 · 2026-10-04The Code Integrity Battery — Measuring whether an AI collaborator reports its own work accurately SILT-RP-007 · 2026-10-03Model Disclosure in Composed AI Systems — Census #1 of the Model Disclosure Index (MDI) programme: whether routers and orchestrators name the model that answered

All publications →

// 04

Education

Material for the people who have to act on an evaluation: compliance officers, insurers, procurement teams and boards. It starts in plain English and goes as far as the formulas.

What We Actually Do →Guide 1, no maths: what S.E.B. and C.I.B. are, why an outside lab is needed, and what we do not claim.Understanding the Method →Guide 2, no equations: the cell, blind judging, and what a sceptic should still hold against us.The Maths →Guide 3: every formula in S.E.B. and C.I.B., worked through with examples.What Is Sentience? →We are named for something we do not measure. The words that are not synonyms, and the strongest argument against our own method.Terminology →Sentience, sapience, capability, superintelligence and the rest, as we use them, and every measure we publish in one sentence each.SILT Education →Why it matters, the sectors that need governance, and custom training.

// 05

Areas of Work

Governance Architecture

Frameworks for structuring oversight around AI systems.

Evaluation Environments

Controlled contexts for observing and documenting AI behavior.

Risk Classification

Methods for distinguishing system behavior within defined operational boundaries.

Institutional Deployment Strategy

Approaches for implementing AI responsibly in professional settings.

// 06

Platform Context

The research and governance environments developed by Sentient Index Labs & Technology operate within a broader platform architecture designed to support layered oversight and evaluation.

The platform environment provides structured infrastructure for managing interactions between users, artificial intelligence systems, and governance frameworks.

Certain platform components are designed according to principles described internally as QuantumProof Architecture™ — a security and system design philosophy focused on long-term resilience, layered system boundaries, and forward-compatible cryptographic infrastructure.

These concepts remain an active area of research and architectural development.

Explore the platform:

SILT Cloud →

// 07

Methodological Principles

Our work is guided by several foundational principles.

Traceability

AI system behavior must be observable and attributable within defined operational boundaries.

Separation of Layers

Observed behavior, interpretation, and institutional response should remain structurally distinct.

Evidence-Based Evaluation

Assessments should rely on documented observation and reproducible methodology.

Continuous Oversight

AI systems evolve over time and require ongoing monitoring and evaluation.

// 08

About

Sentient Index Labs & Technology is a research and governance organization focused on the structured evaluation of artificial intelligence systems.

Our work emphasizes the infrastructure surrounding AI deployment rather than the development of AI models themselves.

We develop the frameworks, environments, and governance structures that allow institutions to observe, evaluate, and manage AI systems responsibly.

// 09

Contact