TRUSTED DATA CATALOG
Find trusted datasets for AI.
WHY A DATASET NEEDS A LABEL
A model is only as good as the data under it. A dataset with no source is a liability: you cannot tell a real figure from a planted one, you cannot reproduce it, and you cannot defend the output it produced. A trusted dataset answers three questions on sight: where did this come from, when was it read, and what has been done to it since.
BROWSE THE CATALOG
Markets
Prediction markets, signed
Live prediction-market prices and resolutions, each observation signed at the moment it was read.
Open the datasetGovernment
US legislation, signed
US bills, resolutions and legislative actions from Congress.gov, each record signed at contact.
Open the datasetGovernment
Federal rulemaking, signed
US rules, proposed rules and agency notices from the Federal Register, each signed at contact.
Open the datasetHazards
Earthquakes, signed
USGS earthquake events (location, magnitude, depth, time), each reading signed at contact.
Open the datasetEconomics
Macroeconomic series, signed
Macroeconomic time series from FRED (rates, prices, output), each value signed at contact.
Open the datasetFinance
SEC filings, signed
Company filings from SEC EDGAR, each filing record signed at the moment of contact.
Open the datasetGovernment
US federal spending, signed
US federal awards and spending from USAspending.gov, each record signed at the moment of contact.
Open the datasetWorld
Global events, signed
World events from GDELT, each event signed at the moment it was read.
Open the datasetSoftware
GitHub activity, signed
Notable GitHub releases and security advisories, each record signed at the moment of contact.
Open the datasetEnvironment
Weather observations, signed
City weather from Open-Meteo (temperature, wind, precipitation), each reading signed at contact.
Open the datasetFinance
Bank financials, signed
US bank financial data from the FDIC, each record signed at the moment of contact.
Open the datasetDecentralized AI
Morpheus network, signed
Providers, models, sessions and staking on the Morpheus AI network, each read signed at contact.
Open the datasetDecentralized AI
Bittensor subnets, signed
Bittensor dTAO subnet data read from the chain, each reading signed at the moment of contact.
Open the dataset00The packs
One service. Every data door.
Data is one thing at DRM3. Every door below serves a slice of the same signed record, and every slice carries a receipt you can verify offline against the published keys.
Partners: pull on your own clock
Every stream pages on a keyset cursor, so a partner pulls what landed since its last cursor on any cadence, nightly included, and never re-reads a row. One DRM3 key meters every pull; where a product sells per call, x402 settles it with no key at all.
What a receipt proves
A signature proves attribution and integrity: this DRM3 key served these rows, and they have not changed since. It never proves the rows are correct. Check any receipt yourself on the verify page, offline.
EVERY DATASET CARRIES A RECEIPT
Each source signs its rows at the moment of contact, on its own key. The receipt names the source, the time, and each step since, and chains to the one before it. A signature proves who recorded the row and that it has not changed; it does not certify the figure is correct, so you get a source to check and a chain to follow, not a claim to take on faith.
Verify a receipt yourselfA DATA COOPERATIVE
The record is pooled and shared. Data comes in signed, and what you pull carries the same label it arrived with, so a dataset keeps its provenance as it moves from one hand to the next. Partners can bring their own sources onto the same rails and offer them under their own key.
Pull a trusted dataset into your agent.