AI Safety And Governance Orchestration
asago is an open-source, community-governed framework that aims to automate the journey from AI governance policy to production-ready, safely deployed AI systems — bridging the gap between compliance teams, AI engineers, and infrastructure operators.
The Problem We're Solving
Organisations today face a months-long manual process to translate their custom AI governance policies into safely deployed AI systems. It involves interpreting policy documents by hand, running disconnected safety tools, manually selecting safety controls such as guardrails, and configuring infrastructure — with no end-to-end traceability.
How It Works
asago aims to connect three things that are currently disconnected: your organisation's AI governance policies, the safety testing of your AI systems, and the deployment of controls to address the risks that testing uncovers. The goal is a workflow that is automated, auditable, and driven by your own policy — not generic checklists — while keeping humans in control and able to review, override, or redirect at every step.
Understand Your Risks
Upload your AI governance policy. asago will read and interpret it automatically, mapping your organisation's requirements to recognised international AI risk frameworks.
Test Your AI System
Automated safety testing will be generated from your specific identified risks and run against your AI system — probing for harmful behaviours relevant to your use case, not just generic benchmarks.
Select the Right Controls
Based on what testing reveals, asago will recommend the appropriate safety controls — with the evidence and rationale behind every selection, ready for auditor review.
Deploy with Confidence
Recommended controls will be packaged into ready-to-deploy configurations for your target infrastructure, eliminating the manual work of translating safety decisions into production systems.
Every step aims to produce a traceable audit trail — from the clause in your policy document to the control deployed in production. The goal is to help compliance teams and auditors build a clear picture of how AI systems have been evaluated and controlled in relation to policy.
Who It's For
Most AI safety tooling serves only one part of the organisation. asago is designed to bridge the full policy-to-production workflow — every stakeholder from compliance officer to platform engineer will be a first-class user.
Compliance & Risk Officers
Will enable you to upload AI governance policies and receive structured risk assessments and audit-ready evidence.
Leadership & CIOs
Will enable tracking of AI governance across your portfolio, demonstrating compliance posture to regulators, and moving from policy decisions to production controls in days.
AI Engineers
Will provide safety tests generated from your organisation's actual policies — not generic benchmarks — enabling automated safety evaluation to be integrated directly into your development workflow.
Platform & Infrastructure Engineers
Will deliver ready-to-deploy safety control configurations — no manual translation of safety recommendations into infrastructure code.
Auditors
Aims to provide a traceable record linking policy clauses to test runs and deployed controls, supporting audit and review workflows.
Built in Collaboration
asago is a collaborative, multi-organisation effort spanning industry and academia. The project brings together expertise in AI safety research, open-source software engineering, enterprise AI deployment, AI governance, and the design of evaluation methodologies for AI systems.
No single organisation has all the answers when it comes to making AI safe and trustworthy at scale. asago is built on the conviction that this problem requires sustained collaboration across the researchers who study AI failure modes, the software engineers who build AI systems, the enterprises that deploy them at scale, and the organisations responsible for governing their use.
Project Goals
asago aims to be the framework the AI safety ecosystem needs — open, interoperable, and free from single-vendor capture.
Goals
- Remain open-source and community-governed — no single vendor controls the direction
- Automate policy-to-production, reducing deployment time from months to days
- Provide full audit traceability from policy through to deployed controls
- Empower non-technical users (compliance, risk) alongside engineers
- Integrate with best-in-class tools where they exist; fill gaps where they don't
Anti-Goals
- Duplicating capabilities that already exist in the ecosystem
- Replacing existing red-teaming or safety control frameworks
- Solving model training or development
- Favouring specific AI vendors or cloud platforms
Get Involved
asago will be open-source under Apache 2.0, community-governed, and open to contributors from vendors, enterprises, academic institutions, and individuals.
Interested in collaborating?
We're looking for organisations and individuals who want to shape the direction of the project — whether as contributors, early adopters, or governance participants. Fill in the form below and we'll be in touch.
Express Interest in Collaborating