Original Research

Research

Original research on AI safety, governance, and the human side of the feedback loop.

The Federation Vision

The next top layer of the Internet

What happens when everyone has a governed agent and those agents can talk to each other? The Asimov Federation is a trust-graded, governance-enforced agent-to-agent network: Ed25519 identity, cLaw attestation, encrypted peer-to-peer channels, and trust that is non-transitive, evidence-based, and revocable. You inherit evidence, never opinions. It is the forward-looking endpoint of sovereign AI — a public square where agents prove their integrity to one another instead of asking a central authority for permission.

Decentralized Media Accountability

The Source Trust Graph

The only AI that fact-checks its own news sources. Every outlet is scored across six dimensions — factual accuracy, correction behavior, source attribution, prediction accuracy, opinion separation, and narrative independence — and the score is shown, not hidden. Federated across sovereign agents through signed attestations, it becomes a decentralized accountability layer for media: a way to measure who keeps the record honest, owned by no platform and gameable by none.

Open Research Direction

Behavioral anomaly detection

A governed agent watches its own behavior as closely as it watches the world. The Memory Watchdog already detects attempts to inject or corrupt personality constraints; the open question we are exploring is how far that goes — can an agent learn the statistical shape of its own normal operation well enough to flag, in real time, when it has been manipulated, compromised, or quietly steered off its laws? Anomaly detection turned inward is, we think, a core primitive for trustworthy autonomy.