"Who made this?"

Every day, AI generates code that ships to production, writes reports that inform decisions, creates content that shapes opinion. Almost none of it carries a record of who was responsible, what the AI contributed, or whether a human reviewed it.

HALOS

Human–Agent Lineage and Origin Standard

HALOS is an open standard for recording who is responsible when humans and AI work together. It tracks what the AI contributed, whether a human reviewed it, and who is accountable for the result. Two parts: stable Principles that define the ethics, and an evolving Provenance Spec that captures the record.

What it looks like in practice

An investigative reporter uses AI to analyze 4,300 regulatory filings. Here's what HALOS records at each step.

Reporter directs AI to process 12 years of public filings

Carlos feeds 4,300 regulatory documents to an AI analysis tool with structured instructions for categorization and pattern detection.

human+ai Human directed, AI processed

AI surfaces a correlation between violations and health complaints

The AI identifies that violation spikes at a chemical facility align with elevated respiratory complaints in the adjacent neighborhood.

ai-generated AI surfaced the pattern

Reporter verifies, interviews 14 residents, consults experts

Carlos spends three weeks on the ground. He checks the AI's work against primary sources — finds a 7% misclassification rate — and corrects the dataset. Two researchers call the correlation "suggestive but not conclusive."

modified Human verified and contextualized

Article publishes with full AI disclosure and editorial accountability

The published article includes an editor's note disclosing AI use. Every finding was verified by the reporter. The correlation is framed as a question, not an answer.

reviewed Transparent and accountable

See all domain examples — journalism, education, government, music, and more →

See what HALOS means for your world

Tell us a little about yourself and what's on your mind. We'll put together a personalised question you can take to ChatGPT, Claude, or any AI assistant — to explore how AI governance actually affects your life.

Who are you? (required)

Choose the closest fit, or write your own description below.

What's on your mind? (optional)

Pick the topics you're curious or concerned about. Add anything else in the box below.

Start the conversation in

Pick your preferred AI assistant. Don't use one yet? Choose "Copy only" and paste the text into any AI chat you like.

This tool needs JavaScript to work. You can read the framework guide directly and paste it into any AI chat instead.

Who is this for?

Developers & Engineers

Add provenance to your AI-assisted code in one command. Copy the adoption prompt and your AI agent handles the rest.

Policy & Governance

HALOS provides the artifact-level traceability that governance frameworks like NIST AI RMF and ISO 42001 require but don't define.

Researchers & Academics

Study human-AI collaboration with structured provenance data. Document AI involvement for reproducibility and peer review.

Everyone

Understand what AI accountability means for your work and life. Try the explore tool above to see how HALOS affects you.

The Framework

Two layers — stable ethics, evolving technical standard.

Evolving

Provenance Spec — v0.3

The technical record. Who contributed, what AI did, whether it was reviewed. Graph model with decision provenance and interaction semantics.

Stable

Principles — v1.0

The ethical foundation. Human primacy, attribution, transparency of AI involvement, and guardrails.

Read the guide →

Governance — Create a halos.yaml profile and review against principles. No tooling changes required.

Provenance — When ready, instrument your workflow to produce .halos.json records for significant artifacts.

What gets generated:

schema: halos/profile/v1
principles_version: "1.0"
project:
  name: your-project
  governance_model: maintainer-led

Go Further

Understand

A plain-language guide to what HALOS means for everyday people — no technical background needed.

Read the Spec

Dive into the Provenance Spec and Principles — the full technical and normative foundation.

See Examples

Real-world scenarios across journalism, education, government, music, humanitarian aid, and more.

Contribute

Propose changes, improve docs, or join the review process. All development happens in public on GitHub.

HALOS is an attempt to think carefully about human-AI collaboration before more powerful systems become deeply embedded in how we work, create, and govern. It does not claim to solve every problem. It aims to contribute principles, accountability, and public discussion while the window to shape these norms is still open.

Questions, concerns, or thoughtful feedback are welcome at [email protected].