The Licence Written in Anger
There is a licence file somewhere in the dependency tree of the artificial intelligence system your organisation runs. You have almost certainly never opened it. It is not a formality, and it is not boilerplate. The terms inside it were written, on purpose, to compel a particular behaviour from whoever ships software built on top of them. The person who wrote those terms is not you, and the conditions they set are now conditions you carry.
That is not a metaphor, and it is not a history lesson yet. It is a procurement fact. A large part of the modern AI stack, the compilers, the core utilities, the model-serving glue, the container base images, rests on components released under copyleft licences whose central clause says the same thing: use this, distribute this, build on this, and you inherit an obligation you cannot switch off by ignoring it. The licence terms inside your AI stack were engineered as a weapon. Understanding who forged that weapon, and against what, is the difference between a board that governs its software supply and a board that merely hopes.
To see why the terms read the way they do, go back to a printer that would not work, in a laboratory in Cambridge, Massachusetts, in 1980.
The printer that would not talk
A Xerox 9700 laser printer arrived at the Massachusetts Institute of Technology Artificial Intelligence Laboratory in 1980, a corporate gift. It was fast, and it jammed constantly. For the laboratory's earlier printer, a programmer named Richard Stallman had written a small hack: when the paper jammed, the software notified everyone waiting on a print job so someone could go and clear it. He wanted to write the same fix for the new machine. He could not. The 9700's software had arrived as a sealed binary, with no source code. There was nothing to read and nothing to change.
[BOOK IMAGE: row 9 - MIT AI Lab sealed printer]
Stallman had been a programmer at the AI Lab since 1971, inside a culture where sharing source was simply how work was done. He learned that a copy of the printer software existed. Robert Sproull, who had worked on it at Xerox PARC and had since moved to Carnegie Mellon, held one. Stallman asked. Sproull refused. He had signed a non-disclosure agreement. Years later Stallman summed up what that first encounter with an NDA had taught him: non-disclosure agreements have victims, and he had just met one.
It is worth being honest, as the book is, about this founding story. It has been polished by retelling, and Stallman himself acknowledged that some of the fine detail is uncertain. The specifics matter less than the pattern, and the pattern is exact: a shared resource enclosed, and the enclosure enforced not by a lock on a door but by a legal instrument. That is the shape of the whole story that follows.
The lesson was reinforced two years later, and this time it was personal. On the sixteenth of March 1982, Stallman's birthday, the company Symbolics changed the terms under which MIT could use improvements to the Lisp Machine, cutting the laboratory out of the collaborative loop it had helped build. Stallman spent close to two years reimplementing Symbolics' improvements on his own, matching them release for release. He was not doing it to win a market. He was doing it because the refusal itself was the thing worth answering.
Building the answer
On the twenty-seventh of September 1983, Stallman posted a message titled "Free Unix!" to the net.unix-wizards and net.usoft newsgroups. He announced GNU, a recursive joke standing for "Gnu's Not Unix", a complete Unix-compatible operating system that he would write and give away. In January 1984 he resigned from MIT, deliberately, so that the university could never claim copyright over the code he was about to produce. He had understood, before almost anyone, that the fight would be decided by who held the legal rights, not by who wrote the cleverest code.
In March 1985 he published the GNU Manifesto in Dr. Dobb's Journal, a philosophical argument for why the users of software deserved specific freedoms. Those freedoms, numbered from zero to three with the numbering settled around 1990, drew a hard binary line. Software that respected all four was free. Software that violated even one was not. Two of the four, the freedom to study how a program works and the freedom to share your improved version, are impossible without access to the source code. The sealed binary in the printer had violated exactly those. The manifesto turned a personal grievance into a definition.
[BOOK IMAGE: row 13 - The four freedoms diagram]
In October 1985 Stallman founded the Free Software Foundation as a tax-exempt non-profit and served as its unsalaried president. The Foundation was the institution; it needed tools. The GNU C Compiler, first released on the twenty-second of March 1987, became, in the book's phrase, the first free software hit. A movement now had a manifesto, a legal home, and production-quality code. What it did not yet have was the mechanism that would make the freedoms stick when the code left Stallman's hands.
The hack inside the law
That mechanism was copyleft, and it is the most elegant piece of legal engineering in the whole history. Copyright exists to let an author restrict copying. Copyleft took that same instrument and inverted its purpose. The copyright holder attaches a condition: you may copy and modify this software, but only if you pass the identical freedom on to everyone who receives your version. The right to restrict is used to guarantee that no one downstream can restrict. Eben Moglen, who became the Foundation's general counsel, described the move as taking a structure built to permit control and using it to guarantee that no one ever can. Lawrence Lessig put it more sharply still: using the master's tools to dismantle the master's house.
[BOOK IMAGE: row 18 - Unix lineage split, copyleft and permissive]
The mechanism was codified in the GNU General Public License version one, published in February 1989. Three clauses carried the weight. Source had to be available. Modifications, once distributed, had to be released under the same terms. And no one was permitted to add restrictions that narrowed the rights already granted. The licence did not ask for good behaviour. It compelled it, and it compelled it automatically, the moment you chose to distribute. This was not a slogan printed on a manifesto. It was a term with teeth, and the teeth close on distribution.
The proof that it was a weapon
Here the argument stops needing Stallman, because the other side proved the point for him. Unix had been born at Bell Telephone Laboratories, and for years it had circulated under terms shaped by an earlier antitrust settlement: distributed to universities at the cost of the media, for teaching, effectively as a gift. Then the law changed. The Modified Final Judgment took effect on the first of January 1984, superseding the 1956 consent decree that had kept AT&T out of the computer business, and it freed the company to treat Unix as what it now legally could be: a commercial product.
[BOOK IMAGE: row 15 - Bell Labs, Unix turned commercial]
The company priced it accordingly. System V Release 1, launched in 1983, carried a commercial source licence at roughly forty-three thousand dollars per installation. Educational institutions, the same departments that had spent a decade building the foundations Unix rested on, paid eight hundred dollars, and even that was now a licence rather than a gift. For the people who had contributed to the commons, the shock was visceral. The same body of code, the same law of copyright, and a fifty-fold gap opened between the price of participation and the price of ownership the instant the legal terms allowed it.
That is the demonstration. A licence is not a passive label on a piece of software. It is a lever, and whoever holds it can move the terms of the whole relationship. AT&T pulled the lever one way, from open to priced. Copyleft was built to pull it the other way, and to weld it in that position so it could not be quietly pulled back.
Two kinds of resistance
Not everyone answered enclosure the same way. The book draws the distinction cleanly. The GNU General Public License is resistance through legal compulsion: you must keep it open. The BSD licence is resistance through renunciation: do what you like, we give up the right to make you share. Both are refusals of enclosure. They simply disagree about method.
The consequence of that disagreement is visible on every device around you. Apple's macOS runs on a foundation that includes a substantial body of BSD-derived code. Microsoft's Windows has shipped BSD-derived networking code since Windows 2000. Neither company was ever obligated to share a single modification back, because the permissive licence asked for nothing in return. The permissive path let downstream actors enclose freely; the copyleft path did not. Same commons, same era, two licences, two opposite futures. The licence chose the outcome.
[BOOK IMAGE: row 19 - Enclosure and resistance timeline]
What a board actually inherits
Bring the two threads together and the pattern is the point. Enclosure and resistance are not sequential chapters. They are parallel tracks, and the licence is the switch between them. Every act of enclosure, from AT&T's pricing to the licensing battles running through cloud and now AI, has met a resistance written in the same medium that enclosed it: a licence.
For a board, this is not a story about heroes. It is a condition on the balance sheet. Copyleft is a live licensing term inside modern AI dependencies, and it is engineered to propagate: it travels with the code, and it activates on distribution. That last sentence is my argument and my extension of the history, not a fact I am reporting from the manuscript, and I want the line between the two kept clear. But the mechanism it rests on is sourced and forty years old, and it has not softened with age. Permissive licences carry the opposite property, and that is not automatically the safer one; they let whoever sits downstream of you enclose what you built. Either way, the licence was chosen by someone else, and it is now a term you inherit and have to account for.
The practical move is unglamorous and overdue. Know which licence governs each material component in your AI supply, the way you would know the jurisdiction of a contract. Treat copyleft obligations as a distribution question your engineering and legal functions answer together, before a product ships, not after a customer asks. Distinguish, in your own inventory, between the components that compel you to share and the components that let others enclose your contribution, because those are different risks and they land on different parts of the business. I ran GPL-licensed infrastructure in production at an early New Zealand internet provider from the late 1990s, so I can say plainly that these terms were operational conditions to work under, not abstractions to admire. The organisations that treated the licence as a governance surface were fine. The ones that treated it as fine print found out, at the worst possible moment, that they had signed something they never read.
The open-source dimension is where the discipline actually lives. The freedoms that Stallman defined are now enforced less by argument than by machinery: the SPDX licence identifier standard that lets a tool name every licence in a codebase without ambiguity, the Open Source Initiative's approved-licence list that decides what the term even covers, the REUSE specification and source-code licence scanners that read a dependency tree the way an auditor reads a ledger, and the licence fields inside a software bill of materials, the SBOM, that carry that provenance downstream to whoever consumes your build. None of this touches New Zealand government procurement policy; it is international, community-run infrastructure. But it is the layer where a board's abstract duty to know what governs its software becomes a concrete, automatable check. If you cannot generate the list of licences in your AI stack, you do not yet govern it.
That same question runs straight up into sovereignty. The licence is the control surface. Whoever sets the terms beneath a stack sets the conditions its operators must meet, and copyleft, through the GNU General Public License version one of February 1989, was built precisely so those conditions could not be quietly re-enclosed once granted. The permissive-versus-copyleft split is the same choice at national scale: BSD-derived code sits inside Apple's macOS and Microsoft's Windows with no obligation to return anything, while copyleft code compels its terms forward wherever it travels. The Modified Final Judgment of the first of January 1984 was the moment the terms under Unix flipped from open to priced by a change in law, not a change in code. The sovereignty question, then, is not only who owns the software a nation depends on. It is which licence governs it, and what that licence compels.
Somewhere in your dependency tree there is a licence file you have never read. Who wrote it, and what did they build it to make you do?
The views expressed in this article are entirely my own, informed by more than 30 years of professional experience in architecture, security, and technology leadership in New Zealand. They do not represent the views of my employer, any government agency, or the New Zealand government. My commentary on legislation and policy is analytical, drawing on publicly available sources and my professional expertise in architecture, security, and AI governance. I follow the Public Service Commissioner's Code of Conduct for the Public Sector and social media guidance.
About the Author: Andreas Hamberger is a New Zealand-based enterprise architect and technology strategist. Over 30 years, he has moved from compiling kernels on a 486 to leading cloud, cyber, and AI transformation programmes across government, banking, transport, and aviation. He founded Yoper Linux, served as a technology specialist for Novell during the Linux Wars, and is the author of "Generative AI: Skynet or Heaven" and "Space Mafia." He can be reached at linux@linux.co.nz. His latest book, Free as in Theft: The Hidden History of Open Source Software (Te Pono Limited, 2026), traces the openness-and-enclosure cycle from the first shared source tapes to the AI licensing wars.
I use AI tools, including Sudowrite, Claude, Perplexity AI, DeepSeek AI, ChatGPT, Grok, Copilot, Openart and Gemini, as deliberate production tools, not ghostwriters. This is consistent with my position: AI amplifies human judgement; it does not replace it. The frameworks, arguments, and editorial decisions in this series are original work. AI accelerated the process. The thinking is mine.
References
[1] Hamberger, A. "Free as in Theft: The Hidden History of Open Source Software." Te Pono Limited, 2026. ISBN 978-0-473-78455-3.
[2] Stallman, R. "The GNU Manifesto." Dr. Dobb's Journal, March 1985.
[3] Free Software Foundation. "GNU General Public License, version 1." February 1989.