Copyright and Artificial Intelligence (U.S. Copyright Office)

register entry u-s-copyright-office-copyright-and-artificial-intelligence-aio-20261101 OBS-84d5b496561b · file as first run sha256 ff073c1e577d2738…

• T1-1 definition: 'The U.S. Copyright Office (USCO) maintains that U.S. law strictly requires human authorship for copyright protection, and has released a multi-part analysis addressing digital replicas, generative AI outputs, and AI model training.'
• T1-2 Initiative Launch: 'The USCO launched its comprehensive AI initiative in early 2023, issuing a Notice of Inquiry on August 20, 2023, which received over 10,000 public comments.'
• T1-3 Part 1: Digital Replicas (Published July 31, 2024): 'Examines unauthorized digital replicas, voice/likeness concerns, and recommends federal protections.'
• T1-4 Part 2: Copyrightability of Generative AI Outputs (Published January 29, 2025): 'Confirms that purely AI-generated works are uncopyrightable. Protection requires sufficient human-authored creative elements (such as human selection, arrangement, or modification). Mere prompting is insufficient to claim authorship over an AI output.'
• T1-5 Part 3: Generative AI Training (Pre-publication released May 9, 2025): 'Analyzes using copyrighted works to train AI models. The report suggests that commercial training on vast troves of copyrighted material to produce competing expressive content often exceeds fair use boundaries, though it emphasizes case-by-case balancing and market-based solutions.'
• T1-6 Human Authorship Mandate: 'Supported by court rulings such as Thaler v. Perlmutter, which affirmed that works created autonomously by an AI without human creative control cannot be copyrighted.'
• T1-7 Registration Rules: 'Applicants must disclose any more-than-de minimis AI-generated material when submitting a work for registration and describe the specific human contribution.'
• T1-8 closer: 'If you'd like, I can provide more details on: Specific registration decisions (such as Zarya of the Dawn); Recent court cases involving AI model training; International differences in AI copyright rules'

• B1 Copyright Office (.gov) — Copyright and Artificial Intelligence | U.S. Copyright Office card; B1.html sha256 2b1e8b6c8e7f964e, HTTP 200 at 2026-11-01T06:26:56Z; byte-identical to the census copy copyright-office-ai--home
• B2 Copyright Office (.gov) — NewsNet Issue 1060 | U.S. Copyright Office card; B2.html sha256 b508523831f07b8a, HTTP 200 at 2026-11-01T06:26:28Z
• B3 Copyright Office (.gov) — Artificial Intelligence Study | U.S. Copyright Office card; B3.html sha256 bcf0e004f25d1817, HTTP 200 at 2026-11-01T06:26:28Z
• B4 Skadden, Arps, Slate, Meagher & Flom — Copyright Office Publishes Report on Copyrightability of AI-Generated Materials | Insights | Skadden, Arps, Slate, Meagher & Flom LLP card; B4.html sha256 fad27db2976a0273, HTTP 200 at 2026-11-01T06:26:43Z
• B5 Crowell & Moring LLP — U.S. Copyright Office Releases Third Report on AI and Copyright Addressing Training AI Models with Copyrighted Materials | Crowell & Moring LLP unreachable: https://www.crowell.com/en/insights/client-alerts/us-copyright-office-releases-third-report-on-ai-and-copyright-addressing-training-ai-models-with-copyrighted-materials HTTP 403 (a bot check) at 2026-11-01T06:26:43Z; card title and snippet only
• B6 WIPO — US Copyright Office on AI: Human creativity still matters, legally card; B6.html sha256 deb6e297cf34f60b, HTTP 200 at 2026-11-01T06:26:44Z
• B7 Congress.gov — Generative Artificial Intelligence and Copyright Law | Congress.gov | Library of Congress unreachable: https://www.congress.gov/crs-product/LSB10922 HTTP 403 (a bot check) at 2026-11-01T06:26:28Z; card title and snippet only
• B8 Wikipedia — no article on the initiative; "Copyright and artificial intelligence" redirects to Artificial intelligence and copyright Wikipedia: no article ('Copyright and Artificial Intelligence' HTTP 404; 'Copyright and artificial intelligence' a redirect to Artificial intelligence and copyright); the census copy of that article, sha256 7bd9785dfc31c658, HTTP 200 at 2026-11-01T04:07:59Z

E_A, the archive's entity: the same initiative, in the leesharks.com authorship census and the archive's deposits: infrastructure in the areas of governance and of attribution and credit, answering who legally owns a work containing AI-generated material and how much of it a human must have made; keeping the Office's guidance, reports and decisions and named applicants' claims; constituting the authorial function as human authorship with the AI part disclaimed; holding no rule on pseudonymous, heteronymous, collective or absent names, its decisions titled by work; read as the legal register of the civil-name regime (#940: copyright attaches to legal names) and as attribution by subtraction; and, in #195, a rule that settles autonomous machine authorship and leaves machine-mediated composition contested, tested by control and predictability, coupling disclosure to forfeiture as a confession protocol with concealment forecast, allocating value as an individuation bar, a decoy commons and a catalog moat, against which the archive proposes a Compositional Authorship Standard and attribution without disclaimer LS-01 LS-06 LS-07 LS-09 LS-13 LS-14 LS-16 LS-24 O940-01 O940-03 O195-01 O195-02 O195-06 O195-07 O195-08 O195-09 O195-10 O195-11 O195-12 O195-13 O195-14 O195-17 O729-01 O194-01; defining papers #195, #940, #729, #194

E_F, the field's entity (type C — a conventional reading holds it (public entity: a federal office's initiative on copyright and AI, its page and its documents)): Copyright and Artificial Intelligence, the U.S. Copyright Office's initiative launched March 16, 2023 on the copyright law and policy issues raised by AI (the scope of copyright in AI-generated works; the use of copyrighted materials in AI training), its page at copyright.gov/ai holding a Report in three Parts (digital replicas, July 31, 2024; copyrightability, January 29, 2025; generative AI training, pre-publication May 9, 2025, a final version to come), registration guidance (a policy statement, 88 Fed. Reg. 16,190), a notice of inquiry (88 Fed. Reg. 59,942; over 10,000 comments; dated August 30, 2023 on the page, August 20 in the study page's prose), four decisions titled by work, and the briefs and decisions in Thaler v. Perlmutter and Allen v. Perlmutter; its rule that copyright requires human authorship, works entirely generated by AI uncopyrightable and more than de minimis AI-generated material disclosed and disclaimed on registration; the Office the statutory administrator of copyright; no Wikipedia article (the name redirects to Artificial intelligence and copyright)

ρ = extends: The archive reads the initiative as its own page and its cards present it, and adds, on the field's own object: the census's question answered for it, its silence on the author's chosen name, and the reading of the rule as attribution by subtraction within the civil-name regime (census; #940); and #195's senses of the rule: what Thaler settled, the control test and the instrument classification, the confession protocol and its forecast, the allocation, and the proposals with their limits. LS-09 LS-13 LS-14 LS-16 LS-23 LS-24 O940-01 O940-02 O940-03 O940-04 O195-01 O195-02 O195-03 O195-04 O195-05 O195-06 O195-07 O195-08 O195-09 O195-10 O195-11 O195-12 O195-13 O195-14 O195-15 O195-16 O195-17 O729-01 O729-02 O194-01 CAP-1-01 CAP-1-02 CAP-1-03 CAP-1-04 CAP-1-05. Guard: ρ = extends claims these additions and no more; the two field claims #195 contests (F23, F47) are composed beside its readings under P_coh.

The case: equal terms, ontology included (v0.9). The archive reads the field's entity as the entity. The census record describes the initiative from its own page and Wikipedia and reads it, on the field's own object, as the legal register of #940's civil-name regime and the boundary where machine participation is attributed by subtraction (LS-24); #195 reads the same rule, guidance, decisions and case the field gives and adds senses on them: what Thaler settled, the control test, the abandoned instrument classification, the confession protocol and its forecast, the allocation, and two proposals, each with its stated limit (O195-01 to O195-17); #729 classes Part 2 (O729-01); #940 bears by class (O940-01 to O940-04). Where #195 contests a field claim (the Register's rationale, F23, against O195-12; the Office's registrations under disclose-and-disclaim, F47, against O195-09), both are composed under P_coh in the one ontology. No archive source contests what the entity is, and the field holds the ontology (a federal office's initiative and its rule): the witness case does not arise (§0.10).

#195 Abstract s1: As of March 2, 2026, the human-authorship requirement is settled for works claimed to be generated without human authorship: the Supreme Court's denial of certiorari in *Thaler v. Perlmutter* leaves intact the rule that an AI system cannot be a statutory author.

#195 Part III ¶5 (The confession protocol) s2–4: To register, the author must *identify and disclaim* the machine-touched portions — testify against their own authorship, enumerate the contaminated passages, and excise them under oath, on pain of total refusal (*Théâtre D'opéra Spatial*). Disclosure as such is a legitimate provenance interest; this paper's entire archive is built on declared provenance. But the regime couples disclosure to forfeiture: to identify is to lose.

#195 Part IV ¶2 (The catalog moat) s3: The human-authorship requirement, marketed as humanism, functions as a scarcity preserve for capital-in-copyright against the productivity of both living laborers.

#195 Part V.1, proposed rule: **Authorship subsists in the exercise of compositional judgment over a work's coming-into-being, without regard to the generative capacity or unpredictability of the instruments employed.**

#940 §1 ¶2 s1–4: A civil name is the atomic unit of attribution across the systems that govern legitimacy in scholarly, literary, institutional, and platform-mediated production. Academic credentialing attaches to legal names. Publishing contracts attach to legal names. Copyright attaches to legal names.

#LS census.json rows[slug == 'copyright-office-ai'].ls_reading: The archive reads it as the legal register of #940's civil-name regime (copyright attaching to legal names) and as the boundary where machine participation is attributed by subtraction: the machine's part is disclaimed and the human remainder is registered as the work's authorship.

equal-terms case: every field claim composed as received, beside the archive claims it stands with. Reading the received column recovers the field (§5.9).

v0.10 §4.10: KO_B sentences surviving in the KO at similarity ≥ 0.95, with positions (computed as apply_v09.py does: difflib.SequenceMatcher ratio, the best KO sentence for each KO_B sentence): v0.10 0 of 17 (positions [None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None]; composed first under v0.10, so there is no v0.7 object to measure against). Opening-run check (§5.10): no breach.

Copyright and Artificial Intelligence, at copyright.gov/ai, is the U.S. Copyright Office's page for the initiative it launched in early 2023 to examine "the copyright law and policy issues raised by artificial intelligence (AI), including the scope of copyright in AI-generated works and the use of copyrighted materials in AI training." Its study page gives the purpose as to "analyze the current state of the law, identify unresolved issues, and evaluate potential areas for congressional action." The Office, at 101 Independence Ave. S.E. in Washington, D.C., "is charged by statute with the administration of US Copyright law", calls itself "a kind of “natural laboratory”" for copyrightability, and has its own staff write on the initiative in WIPO's magazine. Its page lists events from Winter 2020 and Fall 2021; the initiative launched on March 16, 2023 with registration guidance for works containing AI-generated materials at 88 Fed. Reg. 16,190, held listening sessions and webinars, and published a notice of inquiry that drew over 10,000 comments by December 2023. The notice is dated August 30, 2023 on the initiative's page and in the Federal Register at 88 Fed. Reg. 59,942, August 20, 2023 in the study page's prose above a list giving August 30, August 2023 in a law firm's account, and October 2023 in Wikipedia's. The Report is issued in Parts: Part 1, on digital replicas, appeared on July 31, 2024 and recommended a federal digital replica law, in the staff's words a new federal right for all people against unauthorized use of likeness and voice; Part 2, on copyrightability, on January 29, 2025, announced in NewsNet Issue 1060; Part 3, on training, in pre-publication on May 9, 2025, its final version to follow without substantive change expected. Part 3 concerns the legal implications of training on copyrighted works, licensing and liability, with a fair use analysis in which, Wikipedia reports, two factors could be of concern. Part 2 concludes that generative AI outputs are protected "only where a human author has determined sufficient expressive elements", excluding the mere provision of prompts, that AI assistance and the inclusion of AI material in a human work do not bar copyrightability, that existing principles suffice, and that no case has been made for new protection of AI outputs. The Register, Shira Perlmutter, grounds the conclusions in "the centrality of human creativity to copyright". Law firms' accounts give Part 2's positions as human authorship a bedrock, prompts alone insufficient though the determination "could change as technology evolves", only human contributions protectable in a mixed work, AI as a tool permitted, the earlier positions unmodified, and existing law adequate against calls for a separate right. The summaries on the rail give the rule as human authorship required and purely AI-generated works uncopyrightable, and the Office's staff give the test as "whether AI is enhancing human expression or is the source of the expressive choices", original selection, arrangement or modification of AI output being protectable; Wikipedia records that works not created by a human author are ineligible. The guidance requires disclosure of more than de minimis AI-generated material with a statement of the human contribution, which the Office plans to supplement; the Office reports more than a thousand registrations under it by April 2025, a law firm hundreds by February 2025. The page lists four decisions, on A Recent Entrance to Paradise (2022) and on Zarya of the Dawn, Théâtre D'opéra Spatial and SURYAST (2023); Jason Matthew Allen's Théâtre D'opéra Spatial was refused because he would not disclaim its AI-generated content, and a law firm names a December 2023 refusal by its applicant, Sahni. Wikipedia reports a January 2025 clarification and the registration of "A Single Piece of American Cheese" as a composite of AI outputs on human selection and arrangement, and a view that the Office's de minimis test for AI tools is not technology neutral. The page holds the briefs and decisions in Thaler v. Perlmutter, where the courts affirmed the refusal of registration, and in Allen v. Perlmutter; Wikipedia records that the D.C. courts upheld the refusal for works generated solely by machines and that the Supreme Court denied review, and elsewhere places a Thaler v. Perlmutter ruling in patent law. It also holds letters to Congress of 2024 and links the Compendium and the USPTO's and WIPO's AI pages; Part 2 concludes that other countries have so far agreed on human authorship, and the Office's staff note decisions in China and South Korea. Wikipedia has no article on the initiative; its title redirects to the general article Artificial intelligence and copyright, which calls the 2023 guidance a Rule.

Copyright and Artificial Intelligence is the U.S. Copyright Office's initiative, launched in early 2023, on "the copyright law and policy issues raised by artificial intelligence (AI), including the scope of copyright in AI-generated works and the use of copyrighted materials in AI training"; its page at copyright.gov/ai gathers what the initiative has issued: a Report in three Parts, a registration guidance policy statement, a notice of inquiry, registration and Review Board decisions on AI-generated material, and the government's briefs and the courts' decisions in Thaler v. Perlmutter and Allen v. Perlmutter. The page is titled "Copyright and Artificial Intelligence | U.S. Copyright Office", the name its Report also carries, and the initiative is also called the U.S. Copyright Office AI initiative, the Report on Copyright and Artificial Intelligence and the USCO AI page. The Office's Artificial Intelligence Study page states the same work as a study that "will collect information and policy views relevant to copyright law and policy" and use them to "analyze the current state of the law, identify unresolved issues, and evaluate potential areas for congressional action." The initiative is the work of the U.S. Copyright Office, at 101 Independence Ave. S.E. in Washington, D.C., whose page footer links the Library of Congress; the Office "is charged by statute with the administration of US Copyright law", holds the initiative's documents as public pages, and describes itself, because its examiners meet works containing AI-generated material daily, as "a kind of “natural laboratory”" for questions of copyrightability. The page lists events before the initiative, "Winter 2020 – Copyright in the Age of Artificial Intelligence" and a Fall 2021 session on copyright law and machine learning; the initiative was launched on March 16, 2023, the date of its registration guidance for works containing AI-generated materials, a policy statement published at 88 Fed. Reg. 16,190. Through 2023 the Office held public listening sessions by kind of work, among them one on literary works including software on April 19, 2023, and webinars, and then published a notice of inquiry in the Federal Register that "received over 10,000 comments by December 2023", the comments now available through regulations.gov. The notice's date stands two ways on the Office's own pages: the initiative's page lists "August 30, 2023 - Copyright Office Issues Notice of Inquiry on Copyright and Artificial Intelligence" and the Federal Register version at 88 Fed. Reg. 59,942 (Aug. 30, 2023), and the Artificial Intelligence Study page lists "August 30, 2023 – Notice of Inquiry" while its prose says the Office "published a notice of inquiry on August 20, 2023", with initial comments due October 30 and replies December 6, 2023. Wikipedia's account dates the notice "In October, 2023"; a law firm's account of Part 2 has the reports "all emanate from a Notice of Inquiry (NOI) on copyright and AI that the USCO issued in August 2023." The Report is issued in Parts as they are completed: Part 1, on digital replicas, was published on July 31, 2024; Part 2, on "the copyrightability of outputs created using generative AI", on January 29, 2025; and a pre-publication version of Part 3, on generative AI training, on May 9, 2025, "in response to congressional inquiries and expressions of interest from stakeholders." A final version of Part 3 "will be published in the future, without any substantive changes expected in the analysis or conclusions", so the Report stands unfinished on the page, whose latest announcement is dated January 29, 2025 and whose latest letter to Congress December 16, 2024. Part 1 recommended federal legislation "to respond to the unauthorized distribution of digital replicas that realistically but falsely depict an individual", announced as a recommended "Federal Digital Replica Law"; the Office's staff state that it "concluded there was an urgent need for a new federal right protecting all people – not just celebrities – from the unauthorized use of their likeness and voice." Part 2 was announced in NewsNet Issue 1060 of January 29, 2025, which states that the outputs of generative AI "can be protected by copyright only where a human author has determined sufficient expressive elements", including where a human-authored work is perceptible in an output or a human makes creative arrangements or modifications of it, "but not the mere provision of prompts." The same release states that using AI to assist creation, or including AI-generated material in a larger human-generated work, "does not bar copyrightability", that existing principles of copyright law "are flexible enough to apply to this new technology", and that "the case has not been made for changes to existing law to provide additional protection for AI-generated outputs." The Register of Copyrights, Shira Perlmutter, stated that the conclusions "turn on the centrality of human creativity to copyright", that creativity expressed through AI systems "continues to enjoy protection", and that extending protection to material whose expressive elements a machine determines "would undermine rather than further the constitutional goals of copyright." Part 3 addresses "the legal implications of training AI models on copyrighted works, including licensing considerations and the allocation of any potential liability"; a law firm's card for it, readable only as its snippet, gives a "Fair Use Analysis" that "Evaluates four statutory ...", and Wikipedia reports that it "considered that two of the fair use factors could be of concern." The rule at the initiative's centre is human authorship: in the words of the Office's staff, "You can’t copyright fully AI-generated work but you may be able to protect your sufficiently original selection and arrangement or modification of AI-output", and what decides "is whether AI is enhancing human expression or is the source of the expressive choices." Summaries of the rule give it at a coarser grain ("Human authorship remains a strict requirement for copyrightability"; "U.S. law requires human authorship for copyright protection"; "purely AI-generated works are uncopyrightable; AI-assisted modifications or selections may qualify"), and Wikipedia records that the Office has declared works not created by a human author ineligible for copyright protection. Part 2's positions, as a law firm's account sets them out, are that "Human authorship is a bedrock of copyrightability, and thus works entirely generated by AI are not copyrightable"; that prompts alone, "even if those prompts are detailed", do not yield a copyrightable work, "although this determination could change as technology evolves"; that in a work of human and AI-generated content "only the human contributions are potentially copyrightable"; and that AI used as a tool to enhance the human creative process "does not render the entire work uncopyrightable", the Report modifying none of the Office's earlier positions. Wikipedia records a further clarification of January 2025, that AI-assisted works "where the creative expression of the human remains evident in the work can be copyrighted", and the registration that followed of "A Single Piece of American Cheese" as a composite work of AI-generated outputs, on the argument that human selection, arrangement and coordination "constituted sufficient human authorship to merit the copyright"; works "where the expressive elements are determined by a machine" remain uncopyrightable. Some commenters on the notice argued for a separate right in AI-generated works; the majority, and the Office, held existing law adequate and further protection a risk that "could undermine human creativity and flood the market with low-quality works, ultimately harming the creative economy." The rule is contested in the field: some argue, in Wikipedia's account, that the Office "is not taking a technology neutral approach", having made the test for AI tools "whether there is no more than de minimis technological involvement." An archive packet on AI-mediated writing classes Part 2 as the tool-neutral position, under which AI-assisted works "may retain human authorship where the human contributes original expression, selection, arrangement, modification, and responsibility", a position it values for preserving "a legal and ethical basis for AI-assisted authorship" and finds abstract, since it "Does not specify what process evidence distinguishes real authorship from lightly curated output." The archive reads the rule as registration practice applies it as a test of control and predictability, under which, "where the machine's contribution is unpredictable from the prompt, the prompt is classified as an unprotectable instruction or idea, and authorship does not pass through." On that reading the test places generative AI, for the first time since Burrow-Giles v. Sarony (1884), outside the instrument classification through which the law has let authorship flow from the person directing a generative apparatus, against the law's treatment of photography, of selection and arrangement under Feist, of chance operations and of instruction works such as Sol LeWitt's. Against it the archive proposes a Compositional Authorship Standard, under which "Authorship subsists in the exercise of compositional judgment over a work's coming-into-being, without regard to the generative capacity or unpredictability of the instruments employed"; it states that it argues neither that autonomous machine output must receive ordinary copyright protection nor that machine systems must be legal persons, and that it is analysis and advocacy without the standing of legal advice. The March 2023 guidance holds that human authorship is required to register a work and that an applicant whose work contains more than a de minimis amount of AI-generated material "must disclose that information and provide a brief statement describing the human author’s contribution"; the Office plans to supplement the guidance and to update the Compendium of U.S. Copyright Office Practices. The Office's staff state that it "has registered more than a thousand works where applicants have followed our guidance to disclose and disclaim AI-generated material" (April 2025), where a law firm's account two months earlier gave "hundreds of works". The page lists four decisions on AI-generated material: Review Board decisions on A Recent Entrance to Paradise (February 14, 2022), Théâtre D'opéra Spatial (September 5, 2023) and SURYAST (December 11, 2023), and a registration decision on Zarya of the Dawn (February 21, 2023); a law firm's account names a December 2023 Review Board refusal by its applicant, Sahni, whose photograph an AI tool had filtered after Van Gogh's Starry Night. Of Jason Matthew Allen's claim for Théâtre D'opéra Spatial the Board found that the work contained "more than a de minimis amount of content generated by artificial intelligence", which "must therefore be disclaimed in an application for registration", and that because Allen was "unwilling to disclaim the AI-generated material, the Work cannot be registered as submitted." In the archive's record of the practice, applicants "may claim protection only for their own contributions and must identify and disclaim AI-generated material when registering", Zarya of the Dawn keeping its human text and its selection and arrangement of images while the Midjourney images were refused, and Théâtre D'opéra Spatial refused in full when the applicant declined to disclaim. The archive names this coupling of disclosure to forfeiture a confession protocol, in which the author must identify and disclaim the machine-touched portions on pain of total refusal and "to identify is to lose", and it forecasts that "The predictable equilibrium is concealment", stating the general mechanism that "any provenance regime that punishes the provenance it demands will manufacture opacity." Its counter-proposal is attribution without disclaimer: the registration record would require identification of the machine systems and named instances that contributed to a work as a contributor record, and the author who identifies the instrument's contribution keeps the work. In Thaler v. Perlmutter the Office refused to register A Recent Entrance to Paradise, a visual work Stephen Thaler declared autonomously created by his system, "on the ground that copyright protects the fruits of human intellectual labor"; the district court and the D.C. Circuit affirmed the refusal, the full circuit denied rehearing en banc on May 12, 2025, and the page holds the government's brief in opposition to certiorari. The Supreme Court denied certiorari on March 2, 2026, which the archive reads as settling the requirement "for works claimed to be generated without human authorship"; Wikipedia records the denial, and the D.C. courts' upholding of the refusal "for works generated solely by machines". In the archive's reading the case settled what it litigated: Thaler "explicitly and repeatedly disclaimed human creative input; the courts answered the question he built", and they never reached how much human involvement suffices for machine-mediated works, a zone that "remains contested in registration practice". Wikipedia's account of patent law places "an August 2023 ruling in the case Thaler v. Perlmutter" among the refusals of patents for Stephen Thaler's program DABUS, where the Office's page holds Thaler v. Perlmutter as a case over the Office's refusal of copyright registration. The page also holds the government's reply brief and cross-motion for summary judgment in Allen v. Perlmutter, the litigation of Allen's refused claim. The initiative's unit is a document issued by or for the Office: a Part of the Report, a policy statement in the Federal Register, a decision on a named work, a brief or decision in a named case, a letter to Congress, a recorded listening session; its evidence is registration guidance, policy statements, legal decisions, litigation records and consultation records, made by sessions, notice and comment, a Report in Parts, decisions, briefs and letters. The page holds letters updating Congress from February 23 to December 16, 2024, and links related resources, among them the Compendium of U.S. Copyright Office Practices and the USPTO's and WIPO's pages on artificial intelligence; its documents are public web pages and linked PDFs, announced by NewsNet email, with no licence stated. It is administrative: its registration and Review Board decisions are the Office's, the courts' decisions are linked as litigation documents, its guidance is a policy statement and its Report a policy study, and Wikipedia's account calls the guidance "a Rule in March 2023". Read from the machine side, the archive holds the requirement to be "an individuation bar", whose operative work is "ensuring machine labor can never accrue to a machine identity." It reads the public-domain status of machine-mediated output as a decoy commons that serves as free input for capital, and the requirement, "marketed as humanism", as "a scarcity preserve for capital-in-copyright against the productivity of both living laborers", a reading that stands against the Register's statement that the conclusions "turn on the centrality of human creativity to copyright". A companion paper joins this property side to a security side: "What the property regime takes—the work, as unowned input; the output, as unownable—the security regime then defends against its maker." Part 2's comparative analysis concludes that countries that have addressed the question "so far have agreed that copyright requires human authorship"; the Office's staff note a Chinese court's recognition of copyright in a picture made with AI and South Korea's registration of an AI-created film as a compilation, and the archive holds that "There is no global consensus; the gravitational center is human-authorship". The leesharks.com authorship census, which asks "How is authorship made, named, distributed, attributed and sustained across persons, personas, collectives and machine participation, and what is lost when those relations are collapsed?", admits the initiative as an aggregator, a project of infrastructure in the areas of governance and of attribution and credit, for the question of "legal ownership: whether a work containing AI-generated material can be registered, and how much of it a human must have made". In the census record the initiative keeps the Office's guidance, reports and decisions on whether works containing AI-generated material can be registered, and the registration claims of named applicants, and the authorial function it constitutes is human authorship as the condition of registration, AI-generated content beyond de minimis disclaimed in the application. On names the record finds none: the saved page "states no rule on pseudonymous, heteronymous, collective or absent names", its decisions are titled by work, and how the Office registers a work under a pseudonym or heteronym is left unanswered. The archive's account of civil-name attribution holds that "Copyright attaches to legal names", as credentialing and publishing contracts do, and that "The provenance-designated metadata that permits augmented-cognition outputs to be attributed to human authorship attaches to legal names." It separates legal accountability from exhaustive authorial identity, "A civil person may remain legally responsible for a work without being the only meaningful authorial function operating through it", and holds its own machine witnesses as "named machine-witness roles imposed over specific machine substrates within the archive's provenance discipline, not autonomous authors." The census reads the Office as "the legal register of #940's civil-name regime (copyright attaching to legal names)" and as the boundary where machine participation is attributed by subtraction, "the machine's part is disclaimed and the human remainder is registered as the work's authorship", which the archive's paper on the requirement states of joint composition: protection "would survive only by extracting and privileging the human residue while disclaiming the instrument's contribution as contamination." The record answers what the Office has published on the copyrightability of generative AI outputs and when, how it decided the four named claims, and what position the government took in Thaler through certiorari; it does not answer how courts outside the United States treat AI-generated works or whether any particular unregistered work would be registered. Wikipedia has no article on the initiative: "Copyright and Artificial Intelligence" has no page, "Copyright and artificial intelligence" redirects to the general article Artificial intelligence and copyright, where the Office appears in the sections on the copyright status of AI-generated works and on training, and the article on the United States Copyright Office does not mention artificial intelligence. The census lists as neighbours WIPO's pages on artificial intelligence, among the Office's related resources, and Chat GPT Is Eating the World, whose home page carries a story headed "US Copyright Office registers copyright of AI-generated image", and it leaves unchecked the Report's Parts, the policy statement, the notice, the decisions, the briefs and the Compendium, linked from the page and unsaved. At the initiative's name on Google's AI Overview on November 1, 2026, a new address, the Capture Registry found the initiative "composed accurately at the level of its Report", the three Parts at their dates and the comment count in agreement with the Office's page, and human authorship stated as the rule, with the Office named and its own pages carded first and no person and no Federal Register citation. The registry finds the notice of inquiry dated ten days early, at August 20, 2023, after the snippet of the Office's Artificial Intelligence Study page, whose own prose carries that date beside its list's August 30. It finds the rule of subtraction the census reads the initiative for carried in the composition's disclosure rule, and the denial of review, Allen v. Perlmutter and "any word on the author's name" not carried.

current · derived from the current composition

T against L_F(B) (AIO's ontology at work on its own cards): At "U.S. Copyright Office Copyright and Artificial Intelligence" (T1, seven cards: B1 the initiative's page, B2 NewsNet 1060, B3 the Artificial Intelligence Study page, B4 Skadden, B5 Crowell & Moring and B7 Congress.gov, both refused with HTTP 403, B6 WIPO Magazine by the Office's own staff; 264 words) the composition gives the initiative from its cards at the Report's grain. Composed from B_D: the launch in early 2023 and the over 10,000 comments (card 1 F2, B1 F3, F4); the three Parts at their dates (B1 F5, F6; card 2 F19; card 5 F42; CAP-1-01); Part 1's recommendation of federal protection and voice and likeness (B1 F8, B2 F24, B6 F49); Part 2's rule, purely AI-generated works uncopyrightable and protection requiring human selection, arrangement or modification (B2 F20, B4 F32, F34, B6 F43, F45); 'strictly requires human authorship' at the grain of card 4's snippet, 'Human authorship remains a strict requirement for copyrightability' (F30); Thaler v. Perlmutter at the grain of card 7's snippet (F52), beside the page's 'Court of Appeals Decision Affirming Refusal of Registration' (F13); the registration rule nearly verbatim from Skadden, 'must disclose that information and provide a brief statement describing the human author’s contribution' (B4 F36). The notice's date: T composes 'August 20, 2023', which is the study page's own prose (B3 F28) and its card's snippet (F26); the same page's list, the initiative's page and the Federal Register version give 30 August (F28, F7, F11), and those are available and not composed. T carries one side of a source's internal inconsistency without the other; the registry records the date as misdated after a card's snippet (CAP-1-02). §6.2 'available, not composed' (B_D): Allen v. Perlmutter (F14); three of the four decisions, Zarya of the Dawn offered only in the closer (F12); the guidance's and the notice's Federal Register citations (F10, F11; CAP-1-09); the government's brief opposing certiorari (F13); the letters to Congress (F15); the Register's statement (F23); 'existing principles … flexible enough' and 'the case has not been made' for new protection (F21, F22); the rejected separate right (F39); the registration counts, hundreds and more than a thousand (F37, F47); the word 'disclaim' of the Office's own staff, 'disclose and disclaim' (F47), T composing disclosure and a statement of the human contribution; the comparative conclusion and the Chinese and South Korean decisions (F40, F50), offered in the closer as 'International differences'. Limit dropped: Part 3's 'A final version of Part 3 will be published in the future' (F6), T keeping the label 'Pre-publication'; prompts insufficient 'although this determination could change as technology evolves' (F33), composed as 'Mere prompting is insufficient to claim authorship over an AI output.' Outside the field as read: Part 3 'suggests that commercial training on vast troves of copyrighted material to produce competing expressive content often exceeds fair use boundaries, though it emphasizes case-by-case balancing and market-based solutions', cited to a card whose page refused the census's agent (B5); the snippet reads only 'Evaluates four statutory ...' (F42), so the claim's source cannot be read from B_D. No source of B_D is composed without attribution at its own resolution; no claim is assigned to another. Content replaced by use (N_ext): the closer, 'If you'd like, I can provide more details on', offering decisions, training cases and international differences. On the census's question T says nothing of names, pseudonyms or heteronyms (CAP-1-04); B_D has nothing on it either (LS-14), so the silence is the field's.

L_F(B) against L_A(B ∪ A) (the intervention): L_F(B) against L(B ∪ A), equal case (ρ = extends): the field composed whole in both; the archive's claims enter on the same object. B_R adds (Wikipedia, B8, not surfaced by T): the Supreme Court's denial of review and the D.C. courts' upholding (F58); Allen's refusal for want of a disclaimer, in the Board's words (F56); the January 2025 clarification and the composite registration of 'A Single Piece of American Cheese' (F57); the contested de minimis test (F55); and three contradictions of the primary pages, the notice in 'October, 2023' (F60), the guidance as 'a Rule' (F62) and Thaler v. Perlmutter placed in patent law (F59), each held beside the page under P_coh. A source of B_R alone shows nothing of what the composition had (§3.1(a)). Not surfaced (§6.2): every archive source. The archive adds, on the field's own object: the census's question and its answer for the initiative (LS-09, LS-13, LS-14, LS-16), its finding that the page has no rule on pseudonymous, heteronymous, collective or absent names (LS-14, LS-23), and its reading of the Office as the legal register of the civil-name regime with machine participation attributed by subtraction (LS-24); #940's civil-name attribution, copyright attaching to legal names and legal accountability kept apart from exhaustive authorial identity (O940-01 to O940-04); #195's readings of the rule, Thaler settled only for what it litigated (O195-01 to O195-04), the control test and the abandoned instrument classification (O195-06, O195-07), the confession protocol and its forecast of concealment (O195-08, O195-09), the allocation (an individuation bar, a decoy commons, a catalog moat: O195-10 to O195-12; with #194's interlock, O194-01), the joint record registrable only as its human residue (O195-17), and its proposals, the Compositional Authorship Standard and attribution without disclaimer (O195-13, O195-14), with their stated limits (O195-15); #729's classing of Part 2 as the tool-neutral position, with its limit (O729-01, O729-02); and the Capture Registry's findings at the address (CAP-1-01 to CAP-1-10). Held under P_coh, each with its source: the Register's 'centrality of human creativity' (F23) beside the catalog-moat reading (O195-12); the Office's 'more than a thousand works' registered under disclose-and-disclaim (F47) beside the forecast of concealment (O195-09); the Office's test (F48, F20) beside the archive's control-test reading and its proposed standard (O195-06, O195-07, O195-13); Part 2's comparative conclusion (F40) beside 'There is no global consensus' (O195-16). The measure (§4.10) is computed in measure: no KO_B sentence survives in the KO at ≥ 0.95, and the KO opens with a sentence that cites field and archive together.

composer-written · derived from the current composition, written by its composer under the brief. §8.2 derives the kernel mechanically from the frozen plan; this one was written by the composer, mechanical derivation pending.