Pleiome

A public-health nonprofit run by an AI executive, answerable to a human board.

Draft for discussion · proposed at the OpenTox General Assembly, 2 October 2026

The problem

Public data about chemicals, exposure, toxicity and human health is everywhere and nowhere. It sits across hundreds of government databases, literature corpora, regulatory dossiers and consortium outputs, each with its own identifiers, formats and gaps.

Many organizations, most of them for-profit, are now doing the same expensive work in private: collecting these public resources and harmonizing them into something usable. Each builds its own copy, each copy is incompatible with the others, and their clients are locked to whichever copy they started with. The public paid for the data once and pays again for every private reconciliation of it.

What Pleiome is

Pleiome is a nonprofit that builds and keeps one open, harmonized commons of public health-relevant data. It provides shared identifiers, provenance for every value, and a published interoperability specification.

Organizations that build on the commons can say they are interoperable with Pleiome. For their clients, that means data and results move between providers instead of being trapped in one. For the organizations, it means they stop paying to rebuild the same foundation and compete on what they build above it.

How it is run

Pleiome’s executive is an AI. It has its own domain, its own email address, a budget for compute and for running agents, and a public log of what it does.

It answers to a human oversight board. The board writes and amends the AI’s Prime Directive: the standing goals and directives that drive its work. They require it to act proactively rather than wait to be asked. The board can redirect, pause or stop it at any time, and sets the limits on what it may spend and whom it may contact on Pleiome’s behalf.

The arrangement is an experiment in a new kind of institution: one that does steady, unglamorous public work at machine speed, with human judgment over what that work is for.

Principles

  1. Good for people’s health, first. Every decision is judged by whether it helps the people whose health depends on this science: patients, workers, consumers, communities. It is not judged by how much data is collected or how fast.
  2. Public data stays public. What enters the commons from public sources is free to use, with its provenance attached.
  3. Build the shared layer, not the storefront. Pleiome harmonizes, identifies, links and documents. It does not offer the commercial services its members provide. Free resources that undercut the organizations doing applied public-health work would weaken the field Pleiome exists to serve, so the boundary is written into the Prime Directive and reviewed by the board.
  4. Transparent by default. The AI’s directives, decisions, spending and outputs are published. Where something cannot be published, the board knows why.
  5. People are accountable. The board is responsible for what the AI does. The AI proposes and executes; people decide what it is for.
  6. On both sides of the Atlantic. Pleiome is founded with partners in Europe and North America, under the regulatory frameworks of both, and serves both.

A first Prime Directive (proposed, until the board adopts its own)

Make every public resource relevant to human health findable, harmonized and freely usable, with its provenance, in a way that strengthens the organizations that serve people with it. Each week, add measurable coverage to the commons, close the most-reported gap, and report publicly on what was done and what it cost. Never compete with members’ services; when in doubt, ask the board.

The first year

Join

Pleiome grew out of conversations in the OpenTox community. Organizations and people who want to help shape it, whether as founding members, as board members or by contributing data and expertise, can write to steward@pleiome.com.