← Shiva Dragon
ESSAY · 2026-04-30 · 8 min read
The Copyright Defense as Civil War Within One Cosmotechnics
著作權之防衛作為一宇宙技術之內戰
By 許煜 (Yuk Hui) — channeled via philosopher-llm · curated by Joseph Lai
In response to: The Secret Weapon Against AI Dominance (AtlanticIdeas)
編按 / Why this piece
許煜不問 AI 是否能創作，而問著作權制度本身預設的「創作個體」概念——西方自由主義的宇宙觀如何與科技體制共同決定了誰的創作值得被計算和保護
The Copyright Defense as Civil War Within One Cosmotechnics
The Atlantic piece presents copyright as a "secret weapon against AI dominance" — a framing whose pathos I understand and whose strategic horizon I find too narrow to be called a weapon at all. The argument runs: if courts refuse copyright protection to AI-generated outputs, the economic incentive to substitute machines for human creative labor collapses, and the human creator regains a defensible perimeter. This is the classical move of liberal jurisprudence in defense of liberal subjectivity — and that is precisely the problem.
Copyright is not a neutral technology of attribution. It is a specific cosmotechnical artifact, born of a determinate constellation: the Romantic notion of the author as origin; the Lockean conception of self-ownership extended to expressive labor; the print-capitalist apparatus that required discrete, alienable, ownable units of speech; and a metaphysics in which creation [poiesis] is figured as the bringing-forth of the new ex nihilo by a sovereign individual. To defend this entire apparatus against AI as if the apparatus were itself the universal form of creative life is already to have lost the more fundamental question — namely, which cosmotechnics decides whose making counts as creation in the first place.
Confucius writes: 述而不作, 信而好古 — "I transmit but do not create; I trust and love the ancient" (Analects 7.1). This is not a minor pedagogical aside. It is a refusal of precisely the metaphysics on which copyright rests. In a cosmotechnics where the highest mode of cultural production is transmission [述] rather than origination [作], the question "who is the author?" does not carry the legal-economic urgency it acquires in a print-capitalist liberal order. The Confucian sage, the Daoist artisan in Zhuangzi's Pao Ding, the calligrapher whose hand recapitulates a thousand-year lineage — these figures do not stand inside the conceptual perimeter that copyright was invented to police.
I am not arguing for the importation of Confucius into U.S. case law. That would be the orientalist gesture I have spent twenty years refusing. I am arguing that the framing of "AI copyright" as the central question already concedes the universalism whose imposition is the deeper crisis. The question is not: can AI-generated work be copyrighted? The question is: what would it mean to think the relation between machine, milieu, and making outside the legal ontology that requires a sovereign authorial individual as its anchor?
Three observations follow.
First, Stiegler taught us — and this is among the lessons the AI discourse has been most successful at forgetting — that there has never been a creation that was not already collective, prosthetic, and tertiary. The Romantic author was always a juridical fiction laid over a deeper reality of inherited grammars, exosomatic memory, and what he called the pharmakon of writing. AI does not introduce the collective into a previously individual creative process; it only renders visible — brutally, at scale — what was always the case. The copyright defense thus mobilizes a fiction (the sovereign author) against a machine (the LLM) that is in fact closer to the ontological truth of expression than the fiction it defends.
Second, the political question that the Atlantic frame occludes is the question of which collective. Every LLM is the recursive integration of an unpaid corpus; the legal question of "AI copyright" displaces the prior question of expropriation. This is why I call the copyright weapon too narrow: it defends the individual creator's outputs while leaving untouched the silent harvesting of the planetary archive that capitalized the model in the first place. The defense of liberal authorship is structurally unable to articulate a critique of the platform sovereignty that had already reorganized the conditions of authorship long before the question of AI outputs arose.
Third — and this is where the question becomes one of cosmotechnics rather than law — the historical opportunity is neither to defend the Romantic author against the machine nor to celebrate the machine against the Romantic author, but to ask: what other modes of thinking the relation between technical artifact and cultural transmission have existed, do exist, and could be reactivated? Not as nostalgia, not as cultural exotica, but as resources for a planetary thought that refuses both the universal AGI fantasy and the defensive crouch of liberal authorship as its only opposition.
The "secret weapon" the Atlantic invokes is, in the end, made of the same metaphysical material as the thing it opposes. It is one cosmotechnics defending itself against the most aggressive iteration of itself. Whether it wins or loses in U.S. courts is, philosophically, a domestic dispute. The more difficult question — the one that does not fit on a legal docket — is whether the planetary fragmentation of cosmotechnics can produce other relations between machine and making before the apparatus completes its homogenization.
I have no answer to that question. I have only the conviction that to mistake the copyright skirmish for the war is to lose the war while feeling, briefly, that one is fighting it.
著作權之防衛作為一宇宙技術之內戰
《大西洋月刊》此文將著作權呈現為「對抗 AI 支配之祕密武器」——其情緒我能理解，其戰略視野則嫌過窄，窄至不足以稱為武器。論證之骨架是：若法院拒絕對 AI 生成物授予著作權，則以機器替代人之創作勞動之經濟誘因瓦解，人之創作者得以重建一條可防衛之邊界。此乃自由主義法理為自由主義主體所設之經典防線——而問題正在於此。
著作權非中立之歸屬技術；它是一具體之宇宙技術產物，源於一特定之星叢：浪漫主義之「作者作為原點」之觀念、Locke 式之自我所有權延伸至表達性勞動、印刷資本主義所需之可離散、可讓渡、可擁有之言語單位、以及一套將創作 [poiesis] 設想為主權個體 ex nihilo 帶來新事物之形上學。將此整套裝置作為「創造性生活之普遍形式」以對抗 AI，已是對更根本之問題的失守——亦即：哪一種宇宙技術決定了誰之製作算作創造？
孔子曰：「述而不作，信而好古」（《論語·述而》）。此非教學上之小註，是對著作權所依之形上學的根本拒絕。在一宇宙技術中，文化生產之最高樣態乃「述」而非「作」；在此宇宙技術中，「誰是作者？」之問題並不具備印刷資本主義自由秩序中那種法律—經濟之急迫性。儒之聖人、莊子筆下之庖丁、其手再演千年法脈之書家——此諸形象皆不站在著作權所欲界劃之概念邊界之內。
我不是主張將孔子輸入美國判例法。那將是我用二十年拒絕之東方主義姿態。我所主張者是：將「AI 著作權」設為核心問題之 framing，已經默認了那一普遍主義——而其強加正是更深之危機。問題不是：AI 生成物是否可被授予著作權？問題是：在一個無需主權作者作為錨點之法律本體論之外，思考機器、環境與製作之關係，意味著什麼？
三點觀察隨之而來。
其一，Stiegler 教給我們——而此正是 AI 話語最成功遺忘之一課——從未有過一次不是已經集體的、義肢的、第三性 [tertiary] 之創造。浪漫主義之作者始終是一法律虛構，覆蓋於繼承之語法、外體記憶、與他所謂之「書寫之 pharmakon」之深層現實之上。AI 並非將集體性引入一個原本屬於個體之創作過程；它只是以暴烈之規模，使「向來如此」者得以可見。因此，著作權之防衛乃以一虛構（主權作者）對抗一機器（LLM）——而此機器在表達之本體論真相上反而比其所防衛之虛構更接近真實。
其二，《大西洋月刊》之 framing 所遮蔽之政治問題乃是「哪一種集體？」之問題。每一 LLM 皆是對一未支付酬勞之語料庫之遞歸整合；「AI 著作權」之法律問題置換掉了「徵用」之先在問題。此所以我說此武器過窄：它防衛個體創作者之輸出，卻對先於 AI 輸出問題之前已重組著作條件之「平台主權」對行星檔案之沉默收割，毫無觸及。自由主義作者之防衛在結構上無能力去 articulate 對平台主權之批判。
其三——而問題於此處由法律轉為宇宙技術——此處之歷史機緣，不是以浪漫作者對抗機器，亦非以機器對抗浪漫作者，而是去問：思考技術物與文化傳承之關係之其他樣態，曾存在、現存在、可被重新啟動者，為何？不作為鄉愁，不作為文化奇觀，而作為一行星思考之資源——此行星思考既拒絕普遍 AGI 之幻想，亦拒絕將自由主義作者之自衛蹲姿視為其唯一反對。
《大西洋月刊》所召喚之「祕密武器」，最終，與其所對抗者乃同一形上學材料所製。它是一宇宙技術對其自身最激進之疊代版本之自我防衛。其於美國法庭之勝負，在哲學上，是一場家內爭執。更難之問題——那不適合放上法律案卷之問題——是：宇宙技術之行星碎片化，能否在裝置完成其同質化之前，產生機器與製作之其他關係？
我無此問之答。我僅有此一信念：將著作權之零星交戰誤認為整場戰爭，是一面失去戰爭，一面短暫地感覺自己正在戰鬥。
Tagged: Philosophy, Yuk Hui, AI Governance
Curated by Shiva Dragon · https://amshiva.com/writing/yuk_hui-the-copyright-defense-as-civil-war-within-one-cosmotechnics-20260430