Google has introduced an replace to the Open Data Format (OKF), model 0.2. The brand new model provides 5 belief associated options {that a} “client” of the OKF can use to confirm 5 features concerning the OKF bundles.
The 5 belief indicators and their associated OKF fields and idea sort are:
- Provenance (area: sources)
- Belief (fields: generated, verified)
- Freshness (area: stale_after)
- Lifecyle: (area: standing)
- Attestation (new idea sort: Attested Computation)
The announcement positions these 5 indicators as answering 5 questions concerning the OKF bundle in an effort to set up belief.
The 5 questions:
- “What was this created from? (provenance)
- How a lot ought to I belief it? (belief)
- Is it nonetheless true? (freshness)
- Is it the present model? (lifecycle)
- Was this quantity produced the best way we mentioned it have to be? (attestation)”
Provenance
Google introduces a brand new “sources” area that data the place the knowledge in an idea got here from. This permits shoppers to establish the unique sources used to create it. This gives supply info that buyers can use to guage an idea’s trustworthiness, answering t he query, “What was this created from?”.
That is Google’s instance of the sources area:
sources: - id: warehouse-schema useful resource: https://wiki.acme.inside/knowledge/warehouse/schemas/gross sales title: Acme Retail warehouse schema — gross sales dataset creator: workforce:data-platform usage_count: 1240 last_modified: 2026-06-15 - id: revenue-policy useful resource: insurance policies/revenue-recognition.md title: Income Recognition Coverage (FY2026) creator: human:jsmith@acme last_modified: 2026-06-15
Google explains what all of it means:
“The brand new sources area data the supplies an idea derives from: an exterior doc, a bundle-relative path, or perhaps a scope descriptor like “all queries in challenge X.” On the similar time, an entry can carry goal credibility indicators: creator, usage_count, last_modified.
The deliberate selection here’s what we didn’t add. OKF data the indicators, not a credibility rating. A rating is subjective, doesn’t port throughout shoppers, and goes stale the second it’s written.
As an alternative, credibility is inferred from the indicators by whoever is consuming (and could be dynamically scored by the buyer, if desired), the identical method you’d belief a closely used, lately up to date, authoritatively authored supply greater than an nameless one. And when the physique cites a selected supply, it does so with an bizarre markdown footnote keyed to the supply id ([^export-schema]), so attribution is per-claim as a substitute of a dangling record on the backside.”
Belief: Generated And Verified Fields
The following a part of the belief indicators are the Generated and Verified fields.
The “generated” area data who created an idea, whereas the “verified” area data who independently confirmed it. A client (comparable to an AI agent, an LLM, or an utility) can use the verification info to filter ideas based mostly on whether or not they’re unverified, machine-confirmed, or human-reviewed.
That is the instance of the Generated and Verified fields in use:
sort: Metric
title: Income
generated: { by: reference_agent/gemini-2.5-pro, at: 2026-06-30T14:00:00Z }
verified:
- { by: human:jsmith@acme, at: 2026-07-01T09:00:00Z }
Freshness and Lifecycle: standing and stale_after Fields
The standing area communicates what a part of the lifecycle an idea is in so {that a} client can establish whether or not it’s in draft, present, or it’s outdated (deprecated). It indicators whether or not an idea is a draft, steady, or is deprecated.
The stale_after area specifies the date after which the idea needs to be re-verified earlier than getting used. Customers can use these fields to establish ideas that must be re-verified or to exclude outdated ideas from new work whereas preserving them for historic reference.
Right here’s the instance of this in use:
sort: Metric title: Gross Margin (legacy, pre-FY2026) standing: deprecated
Attested Computation: For Verifying Calculations
Attested Computation isn’t a area, it’s a brand new Kind. Attested Computation defines the accredited option to calculate a price and gives a option to confirm that the calculation was carried out the best way it was imagined to be calculated.
The official description explains:
“Provenance solutions the place a declare got here from. Attestation solutions a tougher query that issues the second an agent experiences a greenback determine: was this quantity produced the best way we mentioned it have to be, or did the agent improvise its personal SQL?
OKF v0.2 introduces a brand new idea sort, Attested Computation. It carries not simply what a price means however a sanctioned option to compute it, and the means to test that the sanctioned factor really ran.”
Right here’s the instance:
---
sort: Attested Computation
title: Income for a fiscal 12 months
runtime: bigquery
parameters:
- { title: 12 months, sort: integer, required: true }
executor:
useful resource: expertise/run-on-bq.md
receipt: [job_id, executed_sql, result]
attester:
useful resource: attesters/sql_equality.py
generated: { by: reference_agent/gemini-2.5-pro, at: 2026-06-30T14:00:00Z }
verified:
- { by: human:jsmith@acme, at: 2026-07-01T09:00:00Z }
standing: steady
stale_after: 2026-12-31
sources:
- id: revenue-policy
useful resource: insurance policies/revenue-recognition.md
title: Income Recognition Coverage (FY2026)
creator: human:jsmith@acme
last_modified: 2026-06-15
---
# Computation
SELECT
SUM(
CASE
WHEN o.foreign money = 'USD' THEN o.net_amount
ELSE o.net_amount * fx.rate_to_usd
END
) AS revenue_usd
FROM `acme.gross sales.orders` AS o
LEFT JOIN `acme.finance.fx_daily_rates` AS fx
ON fx.foreign money = o.foreign money
AND fx.rate_date = DATE(o.order_ts)
WHERE o.order_status="delivered"
AND DATE_DIFF(CURRENT_DATE(), DATE(o.order_ts), DAY) >= 30
AND EXTRACT(YEAR FROM o.order_ts) = @12 months"
Up to date Open Data Format Documentation
Google has up to date the GitHub Repository to replicate this replace and has printed an announcement that serves as an explainer.
Featured Picture by Shutterstock/Jack_the_sparow
