GenLayer2026/08/05

Episteme — The Verification Layer for AI Decisions

GenLayerAI VerificationConsensusInfrastructure

Episteme is the verification layer for AI decisions in an open and changing world. It does not ask "What does AI think?" — it asks "Can an AI-generated conclusion survive independent investigation and challenge?" Instead of user → AI → answer, Episteme produces a verifiable decision: evidence-backed, independently challenged, consensus-verified, and versioned.

The Pipeline

Ask → Investigate → Challenge → Verify → Use

1. Ask — a user or application submits a question, a policy, and evidence URLs. 2. Investigate — five agent roles run inside the contract and read live web content: Research, Source quality, Fact-check, Analyst, Skeptic. 3. Aggregate — evidence is scored against the policy's criteria. 4. Challenge — the skeptic hunts for contradictions; anyone can open a public dispute. 5. Verify — GenLayer validators independently re-run the investigation and must reach consensus. 6. Use — a versioned, shareable record with full provenance.

Verifiable Decisions

A decision is an object, not a sentence

Every decision carries provenance: verification ID, decision (PASS/FAIL), confidence, status, consensus, criteria results, sources with category and authority, challenges, revision history, disputes, and an expiry date. It can be challenged, appealed, re-verified, and referenced — not just displayed.

Consensus, the way GenLayer designed it

Implements the Equivalence Principle: a leader proposes a result and validators independently re-run and compare. Decision class must agree, confidence must fall within tolerance, criteria results use direction tolerance (a PASS↔FAIL flip is never accepted), and evidence counters must be consistent. The verdict is genuine majority agreement of independent models.

Built for the Real World

One engine, many verticals

grant-v1

GitHub, GitLab, docs

Open-source grant eligibility

dao-proposal-v1

gov.uniswap.org, snapshot.org

DAO treasury proposals

company-status-v1

registries, official sites

Is an organization still operating?

insurance-claim-v1

flight trackers (required)

Did the claimed event occur?

service-status-v1

live APIs

Is a service live and returning valid data?

Robust by Construction

No party — not even the deployer — can manipulate a result

Immutable policies seeded at deploy, consensus on every mandatory criterion, sources bound to what was actually submitted, original URLs preserved in every revision, tx-returned verification IDs, owner-only re-verification with full revision history, dispute modes (public/parties), evidence provenance, anti-fake-evidence origin constraints, honest failure (NEEDS_REVIEW / INSUFFICIENT_EVIDENCE), expiry, events, and hard input limits.

Measured, Not Claimed

42 direct-mode tests

Including adversarial cases: duplicate register rejected, non-owner re-verify rejected, double dispute rejected, irrelevant evidence → FAIL, allowed_origins enforcement, fabricated-source and fabricated-citation rejection, and mandatory-criterion consensus. Accuracy: 8/8 against ground truth with unanimous consensus. genvm-lint check passes with no forbidden patterns.

Contract

Item	Detail
Network	GenLayer StudioNet
Contract	`0x8934dfd22A3CF8082B443AF80786da2EFE646f08`
Write	`verify_with_policy`, `challenge`, `reverify` (owner-gated)
REST API	`https://episteme-verify.vercel.app/api` — `/verify`, `/challenge`, `/reverify`, `/policies`, `/verifications`

Security & trust model

No admin keys — the deployer has no special power after deploy. The only writes are verify_with_policy, challenge, and owner-gated reverify. No silent rewriting — re-verification appends to revision history; disputes are append-only.

More projects

GenLayer

Aequitas Lex — AI Legal Severity Assessment on GenLayer

GenLayer

Praetor — AI-Powered Escrow Marketplace on GenLayer