Home → We Use AI Responsibly
Transparency about our process

We use AI.
We say so. We explain how.

This platform is honest about something most publications are not: artificial intelligence plays a significant role in how content here is produced. We think that transparency about this is not a weakness — it is part of the same commitment to visible methodology that governs everything else we do.

How it works

A collaboration between
AI and human judgment

No piece on this platform is published by AI alone. No piece is published without AI assistance. The two work together — with clearly defined roles for each.

Artificial Intelligence
Research, drafting & cross-referencing
Initial research and identification of relevant primary sources
Drafting article structures and content for human review
Cross-referencing claims across multiple models independently
Identifying potential gaps, counter-arguments, and limitations
Speed and breadth across large documentation bodies
⇄
Collaboration
Human Editor
Verification, judgment & final authority
Reads primary source documents directly — not AI summaries
Makes all final editorial decisions
Approves every piece before publication
Takes full editorial responsibility for published content
Manages all corrections and disputes
Scope and limits

What AI does —
and what it does not

✓ AI is used for
✓
Initial research and source identification
✓
Drafting article structures for human review
✓
Cross-referencing claims across models
✓
Flagging potential gaps and counter-arguments
✓
Structuring source hierarchies and limitations
✕ AI is not used for
✕
Final editorial decisions — those are human
✕
Source verification — always done by a human reading the original document
✕
Publishing without human review and approval
✕
Replacing primary source reading with AI summaries
✕
Making corrections decisions
Verification process

Three layers of
fact-checking

Every significant factual claim goes through three independent verification layers before publication. No single layer is treated as sufficient on its own.

1
Multi-Model AI Cross-Check AI Layer
The claim is run through Claude, Gemini, and Perplexity independently. Each model’s response is compared. Disagreements between models are flagged for human investigation rather than resolved algorithmically. Agreement across models increases confidence; disagreement signals deeper scrutiny is needed.
2
Human Source Verification Human Layer
A human reviewer goes directly to the primary sources cited — not AI summaries of those sources, but the original documents themselves. UN resolutions, ICJ rulings, OCHA reports, human rights organization findings. The claim is checked against the document, not against a description of it. This layer cannot be delegated to AI.
3
Open Access Confirmation Reader Layer
Every source used must be publicly and freely accessible online. If a source is behind a paywall or requires institutional access, it is not used as primary evidence. This is a deliberate editorial decision — if readers cannot verify a source themselves, we do not cite it as proof. This layer belongs to you, the reader.
Reader verification

You can fact-check us.
That is intentional.

The open access principle

Every source linked in every piece on this platform can be opened, read, and evaluated by anyone with an internet connection. We do not ask you to trust our reading of a document. We show you the document.

This means you can verify every factual claim we make by following the links in the article directly to the primary source. If you find that our reading of a source is incorrect, the corrections process is open. We treat reader verification not as a threat but as part of how this platform maintains its standards.

We are aware that AI models are not neutral and that their training data reflects the same media biases this platform exists to examine. We use AI as a research tool, not as an arbiter of truth. The arbiter of truth on this platform is the primary documentary record — and checking claims against that record is always a human act.

AI provides speed.
Humans provide judgment.
The primary record decides.

The collaboration between AI and human is not a shortcut. It is a way of doing more rigorous work than either could do alone — with the human providing judgment, ethical accountability, and final authority, and AI providing speed, breadth, and the ability to cross-reference across a large body of documentation quickly.

“To document is to resist erasure.”