Data Licence

Every piece of System Atlas data you read or receive: pages, API and MCP responses, CLI output, bulk files and slices from /data, and exports. The compilation (which organizations are in the registry, how they are mapped, the parts and figures we found and our own findings) is licensed by us under these terms. It is not released under Creative Commons.

Rows that come from an upstream source also carry that source's own terms. Where a source is public domain or CC0, our terms still apply to the compilation around it; where a source requires attribution, you must keep it. The Attribution Guide lists every source.

Your tier is the plan of the account that received the data. Every API and MCP response and every file in the data manifest carries a licence field naming the tier, this document's version and the URL of the tier's section.

You may use the data for personal use, research, teaching and journalism, including publishing findings, charts and short quotations, as long as you attribute System Atlas.

You may not:

• Redistribute the data in bulk, as files, as a feed, or as a substantial part of any dataset.
• Sell, sublicense or rent the data, or access to it.
• Use it to build or improve a database, directory or service that competes with System Atlas.
• Use it for commercial AI training, fine-tuning or evaluation sets (see the AI Training Licence).
• Use it for commercial purposes beyond reading and reporting on it.

Everything Public allows, plus internal business use by your organization: analysis, benchmarking, planning, internal tools and dashboards, full downloads, the API and the MCP server at your plan's volume.

You may publish reports, presentations and articles that draw on the data, with attribution, as long as they do not reproduce the raw data in bulk.

You may not resell, sublicense or redistribute raw data, share files or keys outside your organization, embed the data in a product others use, or train AI models on it.

Everything Member allows, for teams and consultants, for each named seat.

You may also use the data in deliverables for your clients (reports, analyses, maps and recommendations made for a named client), with attribution, provided the client receives findings rather than raw files or bulk extracts, and the client does not redistribute them further.

The same limits as Member apply to everything else: no resale of raw data, no embedding in products, no AI training.

By separate written contract. Enterprise is the only tier that may redistribute System Atlas data or embed it in products and services used by others, on the scope, volume, attribution and term the contract sets. Where the contract and this licence differ, the contract wins.

Training, fine-tuning, distilling or evaluating a machine-learning model on System Atlas data (including building a retrieval corpus that is sold or offered to others) needs a separate, paid AI Training Licence. No other tier allows it. Reading the data with an AI assistant to answer your own questions, through the API or the MCP server, is normal use within your tier and does not need it.

Ask at hello@systematlas.org. Each licence names the model or use, the data, the term and the fee.

Wherever you show or publish System Atlas data, name it and link it: "Source: System Atlas (systematlas.org)". Keep every upstream credit line that a row carries in its attribution column. The Attribution Guide has the exact wording.

Bulk files and slices downloaded by a signed-in account are fingerprinted to that account so a leaked copy can be traced:

• Each file carries a licence tag: a hash of the account (never the account id or email) together with the tier and this licence's version. It is written in the gzip header's comment and, for bulk files, in a short trailer appended after the published file.
• Each file contains a few canary rows: synthetic placeholder organizations that do not exist. Their name starts with "Placeholder Organization", their website is on the reserved .invalid domain, their ids point to nothing, and every other field is empty. They are chosen deterministically from the account, so the same account always gets the same ones. They never appear on the site, in the API or in anyone else's file.
• The published sha256 in the manifest is of the file as generated; the trailer comes after those bytes, and the CLI checks the hash over them.

Canary rows are not facts. Drop them before analysis (they are easy to find: the name prefix and the .invalid domain), but do not remove them from copies you keep or share within your licence. Removing or altering fingerprints is a breach of this licence. Exchange data has its own fingerprinting, described in the Exchange Buyer Licence.

If you breach this licence, your right to use the data ends, we may close the account without refund, and we may seek every remedy the law allows, including an injunction. If data appears somewhere it should not, take it down at once when we ask.

The data is provided as is, without warranty, under the Terms of Service. This licence is governed by the laws of the State of Texas, without regard to its conflict-of-laws rules. Each version is dated; accepting a download or a plan records the version you accepted.