ARISU

An applied research lab preserving human authority over
intelligence.

ARISU

[ A ] RISU[ R ] esearches[ I ] ntelligence,[ S ] afety &[ U ] nderstanding.

Researching how advanced intelligence can remain understandable, bounded, and accountable to humanity.

Explore Our Work
Scroll
Layer:01Mission

Capability without surrender.

ARISU researches and builds tools to help organisations govern advanced intelligence, making its use more understandable, bounded and accountable.

001Understandingis not access.002Connectionis not consent.003Intelligenceis not authority.
Layer:02Terminal

Built to be understood.

See our current work and its status. Live and answerable.

ARISUTERMINAL / PUBLIC NODE ONLINE

ARISU terminal waiting to initialize.

Type help for options.

Layer:03Research

Questions before conclusions.

We study the infrastructures and institutions shaping advanced intelligence.

  1. Paper 001 Ongoing

    The Governed Agent

    Agents need explicit contracts. Humans retain the release gate.

  2. Paper 002 Ongoing

    The Many Faces of Safety

    Safety has many meanings. We map them to practical controls.

  3. Paper 003 Ongoing

    Organising Knowledge Systems

    Memory shapes model behaviour. Retrieval should preserve context without inventing it.

  4. Paper 004 Ongoing

    Machine-to-Machine Economies

    Machines will transact directly. Payments still need authority and limits.

  5. Paper 005 Ongoing

    Robots, Data & the Physical World

    Embodied systems need grounded data. Physical action demands harder safety boundaries.

Layer:04Observatory

Evidence before spectacle.

Public instruments for examining evidence and incidents.

001

Research Atlas

Ask the literature and trace every answer back to its sources.

Open Corpus002

Boundary Ledger

Examine documented incidents, vulnerabilities, and near misses.

Public record
Layer:05Projects

Systems that remain answerable.

Governing what advanced agents can know, do, share, and release.

Ongoing

MiniGraph

Contract-oriented infrastructure for bounded agent execution.

Coordinates powerful coding agents through explicit context, bounded authority, independent checks, human gates, and durable evidence. Models can propose effects. Release remains a human decision.

BoundariesAuthorityEvidenceUncertainty
Ongoing

State Terminal

Evidence-backed operational awareness for coding agents.

Brings sessions from popular models into one evidence-backed operational view. Surfaces activity, policy, conflicts, and uncertainty without taking ownership of execution.

Local firstProvenanceAuthoritySupervision
Ongoing

Fieldtrip

Consent-bounded collaboration across people, machines and agents.

Creates temporary rooms where people bring selected agents into a shared repository view. Only bounded presence and overlap signals cross; prompts, file contents, credentials, and control remain local.

Explicit entryBounded presenceLocal controlEphemerality
Layer:06Contact

Research should remain open to challenge.

It takes many mindsto govern intelligence.

contact@arisulab.ai