What AI OSI Core Is

AI OSI Core is a decision evidence standard.

It specifies the minimum information required to reconstruct an AI-assisted decision later, including:

• What decision was made

• Who was accountable for it

• What system was used

• What inputs were relied upon

• What constraints applied

• How long the decision was intended to remain valid

The Core is intentionally small, boring, and strict.

It does not attempt to judge decisions.
It does not enforce behavior.
It does not guarantee correctness or compliance.

It simply ensures that decisions do not disappear into institutional memory.

What AI OSI Core Is Not

AI OSI Core is not:

• An ethics framework

• A safety standard

• A compliance certification

• A governance workflow

• A risk scoring system

• A legal or regulatory service

Recording a decision does not make it good, safe, or lawful.

It makes it inspectable.

The Canonical Specification (v0.1)

The authoritative specification for AI OSI Core is published as a versioned, archived technical document.

AI OSI Core v0.1 → View on Zenodo (DOI)

Version v0.1 defines:

• The canonical Decision Record fields

• Required vs optional elements

• Conformance criteria

• Temporal validity rules

• Scope discipline (what Core explicitly does not define)

Changes to the Core are conservative and rare.

Running Code: A Working Reference Implementation

AI OSI Core is not just a document.

There is a working, local-first reference implementation that anyone can run, inspect, or fork.

GitHub Repository → https://github.com/danielpmadden/ai-osi-core

The repository includes:

• A browser-based Decision Record generator

• Structured validation against the Core schema

• Export to TXT, Markdown, JSON, YAML, and CSV

• No accounts, no telemetry, no network dependency

• Designed to run in secure or air-gapped environments

You can run it locally in minutes.

The App (What It Does)

The AI OSI Core app is intentionally simple.

It helps a user:

• Create a decision record using Core fields

• Understand what information is required (and why)

• Preview the record in multiple formats

• Export and store it as governance evidence

It does not automate judgment, approval, or enforcement.

It exists to make decision memory explicit and durable.

Status

AI OSI Core v0.1 is:

• Public

• Open

• Experimental

• Use-at-your-own-risk

It was not built by a standards body, regulator, or vendor.

It was built to answer a simple question that keeps coming up later:

“Who decided this — and based on what?”

How This Fits the Bigger Picture

AI OSI Core is the foundation of the broader AI OSI Stack.

The Stack addresses layered governance across mandate, data, models, instruction, reasoning, deployment, and disclosure.

The Core stands alone.

You can adopt it without adopting anything else.

Why This Exists

Most AI governance failures are not failures of intent.

They are failures of memory.

AI OSI Core exists to make sure that, years from now, someone can still say:

• This decision happened.

• This is who owned it.

• This is what we relied on at the time.

• This is when it stopped being valid.

Nothing more. Nothing less.