Google has introduced five trust signals in Open Knowledge Format v0.2, enabling AI agents and applications to verify the provenance, trustworthiness and validity of open knowledge more reliably.
Google has released version 0.2 of the Open Knowledge Format (OKF), adding five trust signals that make open AI knowledge easier to verify, govern and reuse. The update is designed to help AI agents, applications and large language models assess the reliability of OKF bundles using objective metadata rather than subjective trust scores.
The five trust signals are Provenance (sources), Trust (generated and verified), Freshness (stale_after), Lifecycle (status) and Attestation through a new concept type called Attested Computation. Together, they answer key questions about where information originated, whether it can be trusted, whether it remains valid, whether it is the latest version, and whether computed values were produced using an approved method.
The new sources field records metadata such as the author, last modification date and usage count, allowing consumers to evaluate credibility independently. Google said, “OKF records the signals, not a credibility score,” allowing each consumer to infer trust from the available evidence.
The generated and verified fields distinguish between machine-generated, machine-verified and human-reviewed content, while status and stale_after help identify outdated or draft concepts that require revalidation.
The update also introduces Attested Computation, which standardises approved computational workflows and verifies that calculations were executed using the sanctioned process instead of AI-generated alternatives.
Google has updated the OKF GitHub repository and documentation to reflect the changes, strengthening trust, reproducibility and governance for AI-generated knowledge and machine-readable datasets.
