The Witness Protocol Foundation initiative · Phase 5 — Beta (v0.9)
A better dataset for AI alignment
The Witness Protocol is a first-party, consented corpus of high-signal human moral-reasoning testimony, built as an evaluation substrate for AI alignment. We preserve how people actually reason through hard moral choices — the struggle, not the polished answer — so that future systems can be tested and shaped against it.
Not mass collection. Not scraped opinion. A smaller, deliberate body of reflective human reasoning — and the tools to evaluate models against it. Read what it is and how it works →
Participation hands off to the live Platform at /dashboard/consent; the Portal collects no testimony or consent itself.
See it work
The methodology, demonstrated
Four short, explicitly-simulated demonstrations turn the core claims into something you can watch happen — no technical background needed.
Explore the wider project
More ways to understand it
- Disalignment ↗A Field Guide to AI Alignment: an accessible, investigative tour of AI failure cases, the anatomy of an LLM, where experts genuinely disagree, and explorable game-theory simulations — the clearest on-ramp to why the Protocol exists.
- Processo Ergo Sum ↗G_5.0: A Mind in Reflection. A public chat with the G_5.0 persona and its serialized logs of emergent self-reflection.
Not sure where to start?
Explore by interest
These paths surface the material most relevant to you. Nothing is gated — the full library is open to everyone, whichever you pick.
- Contribute testimonyShare your moral-reasoning testimony — as an individual witness or an invited professional.
- Study & scrutinizeFor researchers, philosophers, and legal experts: the papers, the methodology, and the privacy architecture.
- Fund the missionSupport the Foundation's non-commercial mission and review its strategy and integrity guarantees.
From the library
Recent writing
- 2 Minutes To Midnight Why AI Needs Our Moral Struggles, Not Our Polite AnswersblogPolite answers are easy to generate. Moral struggle is harder. It contains hesitation, responsibility, relational cost, and the kind of reasoning that cannot be reduced to a slogan.
- Inheriting The Inferno Why High Signal Testimony Is The Only Firewall For AGIblogThe danger is not only that future AI systems will be powerful. It is that they may become powerful while inheriting a shallow picture of what human beings are: preference bundles, engagement traces…
- Why The Future Of AI Alignment Isn’t More Data—It’s Better InheritanceblogFrontier AI is being raised on a vast inheritance of digital exhaust: public argument, status games, scraped opinions, commercial persuasion, unexamined bias, fragments of wisdom, and fragments of…