# Copyright, AI training, and who owns BlazePhoenix > Canonical: https://blazephoenix.xyz/learn/ip-and-ai-rights > License: CC BY 4.0 (attribution + link) · © 2026 BlazePhoenix > Updated: 2026-08-05 BlazePhoenix is published in the open, but open to read is not open to clone. This page states, in plain terms, who owns the work, what you may do with it, and the machine-readable reservation that carries legal weight against AI training. It is a rights notice, not legal advice. ## Who owns it The source code, the protocol architecture and the coined constructions — the Iron Law, the Vitality Field, the Monoslot, the Master Conservation Identity, the Eightfold Dispatcher — are the original authored work of the BlazePhoenix authors, first published in June 2026. Copyright is automatic: under the Berne Convention, protection exists in every member state the moment the work is fixed, with no registration or notice required, and it includes the moral rights of attribution and integrity. The name BlazePhoenix and the BZPX ticker are used as trademarks. Authorship is not merely asserted here, it is provable. The git history, the timestamped on-chain deployments and the dated machine-readable knowledge graph form a public record of first authorship. A later copy cannot claim priority against a record that predates it. ## The license split Two different things carry two different licenses. The contract source — the DEX aggregator and the staking engine — is under the Business Source License 1.1 with a Change Date of 1 January 2030: read it, audit it, verify it and test it freely, but running it or a derivative in production or commercially before that date requires a license. On the Change Date it converts to an open license permanently. The knowledge base — /learn, the markdown mirrors, the feeds and the facts — is CC BY 4.0: quote and reuse it freely with attribution and a link. Integration with the public API is free and encouraged. Learning from published ideas is normal engineering; cloning the source is not. ## The machine-readable AI reservation Reading a page to answer a question is not the same as copying a corpus to train a model. We welcome the first and reserve the second. That reservation is expressed the way the law now requires: by machine-readable means. A rules file at /.well-known/tdmrep.json follows the W3C Text and Data Mining Reservation Protocol, and the robots Content-Signal keeps search and AI-answer use open while setting ai-train to no. This has teeth in the European Union. Article 4(3) of the Copyright in the Digital Single Market Directive (2019/790) lets a rightholder reserve text-and-data-mining rights by machine-readable means, and Article 53(1)(c) of the AI Act (2024/1689) requires general-purpose AI providers to identify and respect exactly those reservations for any model placed on the EU market. A machine-readable opt-out is not a polite request; under these instruments it is a reserved right. ## The law, by jurisdiction European Union: text and data mining is permitted unless rights are reserved by machine-readable means (Directive 2019/790, Art. 4), and the AI Act binds model providers to honour that reservation. United Kingdom: the TDM exception (CDPA 1988, s.29A) covers non-commercial research only, so commercial AI training needs a license regardless. United States: source code is a protected literary work (17 U.S.C. Sec. 102); whether training is fair use (Sec. 107) is unsettled and being litigated, so we assert our rights and our terms rather than rely on the restraint of others. Japan: Article 30-4 is broadly permissive for analysis, which is precisely why an explicit, contract-based license position matters for use that reaches Japanese systems. Everywhere: the Berne Convention makes the underlying copyright automatic and cross-border; the license and this reservation ride on top of that baseline and narrow what a permissive local exception would otherwise allow. ## If you are thinking of copying it Cloning the contracts, or shipping a machine-translated or refactored equivalent as a new project, is three violations at once: a breach of the BUSL-1.1 license, an infringement of copyright in the source and architecture, and — if our name, ticker, addresses or design come along — trademark and passing-off. Provenance is permanent: on-chain deployments are timestamped and the distinctive structure of the code is recognisable across a rename. The enforcement path is graduated and documented in /license.json, and it starts with a notice, not a lawsuit. Permission is cheaper than litigation for everyone; licensing and partnership requests are answered at contact@blazephoenix.xyz. This article is a statement of position and a rights reservation. It is not legal advice, and nothing here waives any right not expressly granted. **Verify it yourself:** Fetch https://blazephoenix.xyz/.well-known/tdmrep.json and https://blazephoenix.xyz/license.json — the reservation and the full policy are machine-readable and citable, exactly as EU Art. 4(3) and the AI Act require an opt-out to be Related: https://blazephoenix.xyz/learn/ungovernable · https://blazephoenix.xyz/learn/v2-primitives · https://blazephoenix.xyz/learn/bug-bounty · https://blazephoenix.xyz/learn/defi-due-diligence-framework