{"course":{"slug":"how-software-changed","title":"How Software Development Changed","url":"https://addaly.com/learn/how-software-changed","level":"intermediate","track":"work"},"lesson":{"slug":"copyright-and-training-unresolved","title":"The disagreement, laid out without a verdict","url":"https://addaly.com/learn/how-software-changed/copyright-and-training-unresolved","position":77,"of":87,"minutes":10,"answer":"Training and output raise separate unresolved questions decided differently under US fair use, the EU's opt-out TDM exception, the UK's research-only exception and regimes with no exception at all.","keyIdea":"Training and output raise separate unresolved questions decided differently under US fair use, the EU's opt-out TDM exception, the UK's research-only exception and regimes with no exception at all, so the only sound planning rule is to avoid depending on a particular ruling."},"body":{"format":"markdown","text":"## Why this lesson refuses to conclude\n\nThe legal status of training on copyrighted material, and of what comes out, is being decided right now in several countries at once, on different statutes, with rulings that have already gone in different directions on different facts.\n\nA course that told you the answer would be wrong somewhere and out of date everywhere. So this lesson does something more useful and less satisfying: it lays out the structure of the disagreement so that you can read the next headline properly, and it gives no legal advice, because it cannot.\n\n## The two positions, stated as their proponents state them\n\n**Training is permitted.** Learning statistical relationships from lawfully accessible material is not reproducing expression. A model holds patterns, not copies. Copyright protects particular expression, not facts, style or technique. Humans learn from copyrighted works without licensing them, and a training process that produces a system rather than a copy is transformative in the sense the doctrine is designed for. Requiring permission for every work would make the technology buildable only by those who already own large corpora.\n\n**Training requires permission.** Building the corpus involves making copies, and copying is the act copyright governs, whatever happens afterwards. The analogy to human learning is rhetorical rather than legal: a person reading a book is not a commercial reproduction at scale. Outputs can substitute for the works they were trained on, which goes to market harm. And where the corpus was obtained from unauthorised sources, the acquisition is a separate wrong from the training.\n\nBoth of these are argued seriously by serious people. If one of them seems obviously correct to you, you are probably holding an assumption the other side does not grant.\n\n> Figure: Two positions, as their proponents state them. Training is permitted: Learning statistical relationships is not reproducing expression; A model holds patterns rather than copies; Copyright protects expression, not facts, style or technique; Requiring permission for every work leaves the field to whoever already owns a corpus. Training requires permission: Building a corpus involves copying, which is the act copyright governs; The comparison with a person reading a book is rhetorical, not legal; Outputs can substitute for the works, which goes to market harm; Where the corpus came from unauthorised sources, that is a separate wrong. Both are argued seriously by serious people. If one looks obviously correct to you, you are probably holding an assumption the other side does not grant, and no jurisdiction has settled it.\n\n## Where jurisdictions actually differ\n\n**United States.** No text-and-data-mining exception; the question runs through fair use, which is a four-factor analysis decided case by case. That structure means rulings can and do diverge on facts. Decisions since 2024 have distinguished between the act of training and the means by which the corpus was acquired, treating those as separable questions. Several major cases are unresolved and appeals are pending.\n\n**European Union.** The 2019 Copyright in the Digital Single Market Directive contains TDM exceptions. Article 3 covers research organisations. Article 4 is broader but lets rightsholders **reserve** their rights in a machine-readable way: an opt-out, whose practical implementation is itself contested. Separately, the AI Act requires providers of general-purpose models to publish a summary of training content.\n\n**United Kingdom.** A narrow TDM exception limited to non-commercial research. A broader exception was proposed in 2022 and withdrawn in 2023 after opposition from rightsholders; consultations have continued without settled legislation.\n\n**Japan.** Article 30-4 of its copyright act is comparatively permissive for use in machine analysis, subject to limits where the use unreasonably prejudices the rightsholder.\n\n**India.** No TDM exception. Fair dealing is an enumerated list of purposes rather than a flexible standard, which makes the analysis different in kind from the US one. The question is open.\n\nThe practical consequence: \"is this legal\" has no global answer, and a company operating in four countries may face four analyses of the same activity.\n\n## The separate question about output\n\nWhether *your* generated code is protectable is a different matter from whether the training was lawful, and people merge them constantly.\n\nThe United States Copyright Office has taken the position that material produced by a machine without human authorship is not registrable, while human-authored contributions to a work containing such material can be. Other offices have reached related conclusions by different routes.\n\nWhy an engineer should care: if a competitor copies a product whose code is largely machine-generated, the question of what you own is live. This affects businesses whose defensibility rests on code rather than on data, operations or customers, and it is one of several reasons that code was rarely the moat it was assumed to be.\n\n## What to do while it is unresolved\n\nThe controls are the ones from the previous lesson, and they are all things you would want anyway.\n\nKeep records. Enable the filters and document that you did. Scan releases. Prefer vendors offering indemnities and read the conditions rather than the announcement. Follow your organisation's existing software licensing process, which exists and has an owner.\n\nAnd apply one planning rule: **do not build a business whose viability depends on a particular outcome in a case that has not been decided.** If a ruling in either direction would end your product, that is a risk to name on the risk register, not a question to resolve by picking the answer you prefer.\n\n## What this lesson is not\n\nIt is not legal advice, and nothing in it is a substitute for a lawyer in your jurisdiction who knows your facts. Two organisations doing superficially identical things can get different answers, because the answer turns on details (what was used, where, under which contract, for what purpose) that a course cannot know.\n\n> Anyone who tells you this is settled is describing their preference. The useful skill is holding an unresolved question without either freezing or pretending it resolved in your favour."},"check":{"question":"Why can two companies doing apparently identical training activities face genuinely different legal analyses?","options":[{"text":"Enforcement priorities differ between regulators, so the same activity is pursued in one country and ignored in another.","correct":false,"tempting":"Confuses whether a rule is enforced with whether the underlying rule differs."},{"text":"Larger companies attract litigation, so risk scales with visibility rather than with the activity itself.","correct":false,"tempting":"Describes who gets sued rather than how the legal question is actually decided."},{"text":"Copyright applies only where a work is registered, so coverage depends on which works each company happened to use.","correct":false,"tempting":"Registration affects remedies in some systems but is not what creates protection."},{"text":"The governing rules differ by country: open fair use, an opt-out exception, or none at all.","correct":true}]},"prev":{"slug":"licences-and-familiar-code","title":"When the output looks like something else","url":"https://addaly.com/learn/how-software-changed/licences-and-familiar-code"},"next":{"slug":"maintainers-and-the-flood","title":"The commons under a flood of free submissions","url":"https://addaly.com/learn/how-software-changed/maintainers-and-the-flood"},"authorship":"Founder and series editor: Dr Mustafa Mun. Written by AI models to a written standard. Exam keys checked blind by a second model: 3,665 of 3,666 agreed. Academic review invited.","review":{"status":"Draft, open for academic review"},"cite":"Addaly (2026) 'The disagreement, laid out without a verdict', in How Software Development Changed. Edited by M. Mun. Addaly. Available at: https://addaly.com/learn/how-software-changed/copyright-and-training-unresolved (Accessed: 13 October 2026).","licence":{"name":"CC BY-NC-SA 4.0","url":"https://creativecommons.org/licenses/by-nc-sa/4.0/"},"reader":{"schema":1,"title":"The disagreement, laid out without a verdict","courseTitle":"How Software Development Changed","module":{"index":8,"slug":"provenance-security-and-law","title":"Provenance, security and the unsettled law"},"answer":null,"keyIdea":"Training and output raise separate unresolved questions decided differently under US fair use, the EU’s opt-out TDM exception, the UK’s research-only exception and regimes with no exception at all, so the only sound planning rule is to avoid depending on a particular ruling.","hero":{"kind":"figure","figure":{"t":"figure","label":"Figure","figure":{"left":{"title":"Training is permitted","items":["Learning statistical relationships is not reproducing expression","A model holds patterns rather than copies","Copyright protects expression, not facts, style or technique","Requiring permission for every work leaves the field to whoever already owns a corpus"]},"type":"compare","right":{"title":"Training requires permission","items":["Building a corpus involves copying, which is the act copyright governs","The comparison with a person reading a book is rhetorical, not legal","Outputs can substitute for the works, which goes to market harm","Where the corpus came from unauthorised sources, that is a separate wrong"]},"title":"Two positions, as their proponents state them","caption":"Both are argued seriously by serious people. If one looks obviously correct to you, you are probably holding an assumption the other side does not grant, and no jurisdiction has settled it."},"alt":"Figure: Two positions, as their proponents state them. Training is permitted: Learning statistical relationships is not reproducing expression; A model holds patterns rather than copies; Copyright protects expression, not facts, style or technique; Requiring permission for every work leaves the field to whoever already owns a corpus. Training requires permission: Building a corpus involves copying, which is the act copyright governs; The comparison with a person reading a book is rhetorical, not legal; Outputs can substitute for the works, which goes to market harm; Where the corpus came from unauthorised sources, that is a separate wrong. Both are argued seriously by serious people. If one looks obviously correct to you, you are probably holding an assumption the other side does not grant, and no jurisdiction has settled it."}},"blocks":[{"t":"h2","id":"why-this-lesson-refuses-to-conclude","runs":[{"text":"Why this lesson refuses to conclude"}]},{"t":"p","runs":[{"text":"The legal status of training on copyrighted material, and of what comes out, is being decided right now in several countries at once, on different statutes, with rulings that have already gone in different directions on different facts."}]},{"t":"p","runs":[{"text":"A course that told you the answer would be wrong somewhere and out of date everywhere. So this lesson does something more useful and less satisfying: it lays out the structure of the disagreement so that you can read the next headline properly, and it gives no legal advice, because it cannot."}]},{"t":"h2","id":"the-two-positions-stated-as-their-proponents-state-them","runs":[{"text":"The two positions, stated as their proponents state them"}]},{"t":"p","runs":[{"text":"Training is permitted.","b":true},{"text":" Learning statistical relationships from lawfully accessible material is not reproducing expression. A model holds patterns, not copies. Copyright protects particular expression, not facts, style or technique. Humans learn from copyrighted works without licensing them, and a training process that produces a system rather than a copy is transformative in the sense the doctrine is designed for. Requiring permission for every work would make the technology buildable only by those who already own large corpora."}]},{"t":"p","runs":[{"text":"Training requires permission.","b":true},{"text":" Building the corpus involves making copies, and copying is the act copyright governs, whatever happens afterwards. The analogy to human learning is rhetorical rather than legal: a person reading a book is not a commercial reproduction at scale. Outputs can substitute for the works they were trained on, which goes to market harm. And where the corpus was obtained from unauthorised sources, the acquisition is a separate wrong from the training."}]},{"t":"p","runs":[{"text":"Both of these are argued seriously by serious people. If one of them seems obviously correct to you, you are probably holding an assumption the other side does not grant."}]},{"t":"h2","id":"where-jurisdictions-actually-differ","runs":[{"text":"Where jurisdictions actually differ"}]},{"t":"p","runs":[{"text":"United States.","b":true},{"text":" No text-and-data-mining exception; the question runs through fair use, which is a four-factor analysis decided case by case. That structure means rulings can and do diverge on facts. Decisions since 2024 have distinguished between the act of training and the means by which the corpus was acquired, treating those as separable questions. Several major cases are unresolved and appeals are pending."}]},{"t":"p","runs":[{"text":"European Union.","b":true},{"text":" The 2019 Copyright in the Digital Single Market Directive contains TDM exceptions. Article 3 covers research organisations. Article 4 is broader but lets rightsholders "},{"text":"reserve","b":true},{"text":" their rights in a machine-readable way: an opt-out, whose practical implementation is itself contested. Separately, the AI Act requires providers of general-purpose models to publish a summary of training content."}]},{"t":"p","runs":[{"text":"United Kingdom.","b":true},{"text":" A narrow TDM exception limited to non-commercial research. A broader exception was proposed in 2022 and withdrawn in 2023 after opposition from rightsholders; consultations have continued without settled legislation."}]},{"t":"p","runs":[{"text":"Japan.","b":true},{"text":" Article 30-4 of its copyright act is comparatively permissive for use in machine analysis, subject to limits where the use unreasonably prejudices the rightsholder."}]},{"t":"p","runs":[{"text":"India.","b":true},{"text":" No TDM exception. Fair dealing is an enumerated list of purposes rather than a flexible standard, which makes the analysis different in kind from the US one. The question is open."}]},{"t":"p","runs":[{"text":"The practical consequence: “is this legal” has no global answer, and a company operating in four countries may face four analyses of the same activity."}]},{"t":"h2","id":"the-separate-question-about-output","runs":[{"text":"The separate question about output"}]},{"t":"p","runs":[{"text":"Whether "},{"text":"your","i":true},{"text":" generated code is protectable is a different matter from whether the training was lawful, and people merge them constantly."}]},{"t":"p","runs":[{"text":"The United States Copyright Office has taken the position that material produced by a machine without human authorship is not registrable, while human-authored contributions to a work containing such material can be. Other offices have reached related conclusions by different routes."}]},{"t":"p","runs":[{"text":"Why an engineer should care: if a competitor copies a product whose code is largely machine-generated, the question of what you own is live. This affects businesses whose defensibility rests on code rather than on data, operations or customers, and it is one of several reasons that code was rarely the moat it was assumed to be."}]},{"t":"h2","id":"what-to-do-while-it-is-unresolved","runs":[{"text":"What to do while it is unresolved"}]},{"t":"p","runs":[{"text":"The controls are the ones from the previous lesson, and they are all things you would want anyway."}]},{"t":"p","runs":[{"text":"Keep records. Enable the filters and document that you did. Scan releases. Prefer vendors offering indemnities and read the conditions rather than the announcement. Follow your organisation’s existing software licensing process, which exists and has an owner."}]},{"t":"p","runs":[{"text":"And apply one planning rule: "},{"text":"do not build a business whose viability depends on a particular outcome in a case that has not been decided.","b":true},{"text":" If a ruling in either direction would end your product, that is a risk to name on the risk register, not a question to resolve by picking the answer you prefer."}]},{"t":"h2","id":"what-this-lesson-is-not","runs":[{"text":"What this lesson is not"}]},{"t":"p","runs":[{"text":"It is not legal advice, and nothing in it is a substitute for a lawyer in your jurisdiction who knows your facts. Two organisations doing superficially identical things can get different answers, because the answer turns on details (what was used, where, under which contract, for what purpose) that a course cannot know."}]},{"t":"quote","runs":[{"text":"Anyone who tells you this is settled is describing their preference. The useful skill is holding an unresolved question without either freezing or pretending it resolved in your favour."}]}],"check":{"question":[{"text":"Why can two companies doing apparently identical training activities face genuinely different legal analyses?"}],"options":[{"text":[{"text":"Enforcement priorities differ between regulators, so the same activity is pursued in one country and ignored in another."}],"correct":false,"tempting":[{"text":"Confuses whether a rule is enforced with whether the underlying rule differs."}]},{"text":[{"text":"Larger companies attract litigation, so risk scales with visibility rather than with the activity itself."}],"correct":false,"tempting":[{"text":"Describes who gets sued rather than how the legal question is actually decided."}]},{"text":[{"text":"Copyright applies only where a work is registered, so coverage depends on which works each company happened to use."}],"correct":false,"tempting":[{"text":"Registration affects remedies in some systems but is not what creates protection."}]},{"text":[{"text":"The governing rules differ by country: open fair use, an opt-out exception, or none at all."}],"correct":true,"tempting":null}],"order":[1,0,2,3],"why":null},"prev":{"slug":"licences-and-familiar-code","title":"When the output looks like something else","minutes":9},"next":{"slug":"maintainers-and-the-flood","title":"The commons under a flood of free submissions","minutes":9}}}