Trustoracle FAQs

Question 1

What kind of cryptographic evidence does Trustoracle use?

Accepted Answer

Trustoracle utilizes robust cryptographic evidence, including ES256K signing (RFC 7515) for every response, provides JWKS public keys for verification, and can anchor facts on-chain (Polygon/XRPL) to create immutable, tamper-proof timestamped proof of data.

Question 2

How does Trustoracle guide AI agents on decision-making confidence?

Accepted Answer

Trustoracle provides a standardized verdict schema with a confidence score and four action levels: `proceed` (≥85% confidence), `caution` (60-84%), `abstain` (30-59%), and `escalate` (<30%), recommending the appropriate action based on the verified evidence.

Question 3

What is Trustoracle and why is it important for AI agents?

Accepted Answer

Trustoracle provides a verifiable trust layer for AI agents. It's crucial because it allows agents to check claims against cryptographic evidence for provenance, freshness, and integrity before taking action, mitigating risks from ungrounded or hallucinated outputs, especially in regulated contexts.

Question 4

How does Trustoracle verify AI agent claims?

Accepted Answer

Trustoracle verifies claims by comparing them against FeedOracle's ES256K-signed, machine-verifiable evidence. This process includes tracing full data provenance from source to agent, ensuring data freshness with configurable checks, and detecting potential AI hallucinations.

Question 5

Can Trustoracle help detect AI hallucinations?

Accepted Answer

Yes, Trustoracle features a dedicated `hallucination_check` tool. It compares an agent's statement against established ground truth and scans for red-flag patterns (e.g., absolute language, factually suspect claims) to identify and flag unverified or incorrect information.

Trustoracle

Trustoracle

主な機能

ユースケース

主な機能

ユースケース