CMS AI and Copyright Case Tracker

register entry cms-ai-and-copyright-case-tracker-aio-20261029 OBS-fd62b2209523 · file as first run sha256 09dff1bb3bec0de4…

• T1-1 definition: 'monitors generative AI copyright litigation and regulatory developments across Europe, Africa, Asia-Pacific, the Americas, and the Middle East, covering over 40 U.S. claims and numerous international lawsuits' (marker: card 1)
• T1-2 LLM Training & Infringement: 'Monitors litigation addressing whether large language model (LLM) training and text-and-data mining (TDM) constitute copyright infringement (such as Getty Images v. Stability AI).'
• T1-3 Transient & Copying Exceptions: 'Analyzes European and UK cases dealing with transient training copies and machine-readable TDM opt-outs (such as Robert Kneschke v. LAION e.V.).'
• T1-4 AI Authorship Definitions: 'Tracks legal rulings on whether AI-generated outputs can be copyrighted and the requirement for human/natural person authorship (such as cases in the Czech Republic like S. Š. v. TAUBEL LEGAL).'
• T1-5 Global Precedents: 'Covers major international rulings, including the Delhi High Court interim order in ANI Media Pvt Ltd v. OpenAI OPCO LLC, which evaluated AI training under fair dealing provisions.' (marker: card 2)
• T1-6 U.S. & Cross-Border Claims: 'Follows prominent actions involving major media publishers, authors, and tech providers (including Disney, NBC Universal, The New York Times, Dow Jones, and Anthropic).'
• T1-7 closer: 'Would you like information on a specific jurisdiction (such as the UK, EU, or US), or details on a particular case from the CMS tracker?'

• B1 CMS.law — Artificial Intelligence and Copyright Case Tracker card; HTTP 200 at 2026-10-29T05:07:02Z; B1.html sha256 4d331b871f62c2b8; B1.txt sha256 8a42ab193117ddf4
• B2 CMS.law — AI and Copyright Case Tracker in India| CMS card; HTTP 200 at 2026-10-29T05:07:24Z; B2.html sha256 f44c6af992e79cfb; B2.txt sha256 f32620a445a196d0
• B3 CMS.law — AI and Copyright Case Tracker in United Kingdom - CMS.law card; HTTP 200 at 2026-10-29T05:07:26Z; B3.html sha256 b4cea0f1e0d3db49; B3.txt sha256 de566937fb2b5959
• B4 Taylor Wessing — AI & Copyright Case Tracker | Taylor Wessing card; HTTP 403 at 2026-10-29T05:07:53Z; B4-taylorwessing-403.html sha256 b83d6b4335f871d6; B4-taylorwessing-403.txt sha256 cf68e224d2e5d7da; unreachable, title and snippet only
• W1 Wikipedia — CMS AI and Copyright Case Tracker (no article) wikipedia; HTTP 404 at 2026-10-29T05:08:32Z; W1.html sha256 77b41706e3bb3eb4; W1.txt sha256 52cdbc66843e71a6; W1-search.html sha256 8bf8c6cc14d6c2fb; W1-search.txt sha256 1558b87513ceb86b
• W2 Wikipedia — CMS (law firm) wikipedia; HTTP 200 at 2026-10-29T05:08:36Z; W2.wiki sha256 9a20f4fb547b2f8c; W2.html sha256 233ba333daafa4d5; W2.txt sha256 0e33211eb0b181f6

E_A, the archive's entity: the same tracker, in the MMRS meta-aggregator's record: a database compiled by the law firm CMS summarizing 23 court cases on AI and copyright from 10 jurisdictions, led by Germany, newest 31 July 2026, no US cases listed, the page's US count kept apart from the list and its region menu read as navigation to the firm's offices; the legal reception of the provenance question, what courts decide about training on and outputting protected work, beside US litigation trackers the census has not read; observed by the Capture Registry at its name, where the composition layer read the menu as scope and the US count as holdings and named the cases right MMRS-01 MMRS-02 MMRS-15 MMRS-16 MMRS-06 MMRS-07 MMRS-23 CAP-1-01 CAP-1-03 CAP-1-07; defining papers

E_F, the field's entity (type C — a conventional reading holds it (public entity: a database, a law firm's publication)): a publication of the law firm CMS, dated 31 Oct 2025 and 'updated regularly': a copyright + AI case tracker of case summaries, each giving the litigation, each party's arguments and the judgment, in 23 pages from ten jurisdictions led by Germany, with no US case; its executive summary frames the questions its cases turn on and states that over 40 copyright claims are being litigated in the US; Wikipedia has no article for it (B1–B4, W1, W2)

ρ = extends: the archive reads E_F as the entity and adds distinctions and senses on the field's own object: the list kept apart from the page's sentence on US litigation and from the site's menu, the tracker's place as the legal reception of provenance, the neighbour set apart, the record kept apart from the entity and the lifecycle from the observation, the census's own limit, and the composition layer's observed treatment of the entity MMRS-15 MMRS-16 MMRS-06 MMRS-07 MMRS-17 MMRS-23 MMRS-22 CAP-1-01 CAP-1-03 CAP-1-07. Guard: ρ = extends claims an addition and no more: MMRS-06 and MMRS-21 are the record's readings (basis 'reading'), MMRS-07 is the census's limit, not a finding about the tracker; MMRS-15 and MMRS-16 draw distinctions the field's texts allow; where an archive claim meets a field claim (MMRS-16 beside F17; MMRS-10 beside F4) both are composed under P_coh in the one claim space

The case: equal terms, ontology included (v0.9). The MMRS record holds the field's entity as the entity: it describes the tracker from the tracker's own index page, read 2026-10-28 for all fields, with a basis per field (operator, observed, reading), and composes nothing that denies a field presupposition. What it adds is on the field's object: the distinction between the list and the page's sentence about US litigation, between the list and the site's region menu, the tracker's place in the field as the legal reception of the provenance question, the neighbour set apart as a candidate, the record kept apart from the entity and the lifecycle from the observation, and the census's own limit (the US trackers unread); the Capture Registry adds the composition layer's observation at the tracker's name. The ontology the archive uses is available in the field itself: the page sets its list ('23 Pages', each entry by jurisdiction) apart from its executive summary, and its summary speaks of US litigation as of the field ('There are now over 40 copyright claims being litigated in the US'); the archive draws the distinction the field's texts allow. The law of the suppressed witness does not apply (§0.10: not where the field holds the ontology). Not mixed.

#MMRS article, kernel (articles.py): The CMS AI and Copyright Case Tracker, compiled by the law firm CMS, summarizes 23 court cases on AI and copyright from 10 jurisdictions, led by Germany

#MMRS census.json, schema_note: each page has two nodes, the MMRS record and the entity it describes; relations are typed; each assertion carries a source and a basis (operator statement, third-party report, our observation, our reading); lifecycle is kept apart from the observation that saw it

#MMRS census.json, rows[cms-ai-copyright-tracker].note: the page's summary states that 'there are now over 40 copyright claims being litigated in the US'; none is in the tracked list

#MMRS article, Place in the field ¶1 s1: It records the legal reception of the provenance question: what courts decide about training on, and outputting, protected work.

#capture cms-ai-and-copyright-case-tracker-aio-20261029, findings[1]: THE COUNT MADE THE HOLDINGS.

#B1 Executive summary ¶7 s1 (field): There are now over 40 copyright claims being litigated in the US

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).

v0.9 §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): v0.9 0 of 11 (positions [None, None, None, None, None, None, None, None, None, None, None]; composed first under v0.9, so there is no v0.7 object to measure against). Opening-run check (§5.10): no breach.

The Artificial Intelligence and Copyright Case Tracker is a publication of CMS, dated 31 Oct 2025, which its makers describe as a copyright + AI case tracker of accessible case summaries, updated regularly, covering the most topical AI and copyright case law, built with CMS colleagues across the UK, Europe and beyond. Each case summary gives the litigation, each party's arguments and the final judgment, and the case pages carry court, country, parties, the date of the claim and its type. The publication has 23 pages, listed newest first from GEMA v. Suno (31 July 2026) to S. Š. v TAUBEL LEGAL (11 October 2023), from ten jurisdictions led by Germany, with no US case. Its contacts are Sarah Wright and Ben Hitchens in London and, for India, Bharadwaj Jaishankar of CMS Induslaw; cases not yet summarised may be sent by email. Its executive summary sets out new issues of ownership, subsistence and infringement, the reinterpretation of copyright principles, three questions tied to Getty v Stability AI, Kneschke v LAION and the Czech authorship case, and an area likely to grow more complicated with the EU AI Act. It says that US courts have led the way, that over 40 copyright claims are being litigated in the US, and that the first US decision, in June 2025, found training Claude on books non-infringing while sending the pirated materials to trial. The site's region menu lists International, Europe, Africa, Asia-Pacific, the Americas and the Middle East, the Americas including the United States of America, and the page points to the firm's Digital Regulation Tracker. In ANI v OpenAI, a claim issued on 19 November 2024 and summarised on a page of 25 Jun 2025, the Delhi High Court's interim order of 24 July 2026 refused an injunction, holding training use fair dealing at the prima facie stage under a three-factor test and finding that the responses came from retrieval, not memorisation; ANI had claimed that ChatGPT cites it as the source of false information; the main suit continues. In Getty v Stability AI, summarised on a page of 27 Sep 2024, the High Court on 4 November 2025 rejected the secondary copyright claim, the model embodying no infringing copies and its weights encoding no image, after Getty dropped its primary claims; Getty won narrow trade mark issues, the firm reads the judgment as significantly limiting developers' liability, and the page calls Stable Diffusion a text-to-image model. Taylor Wessing's AI & Copyright Case Tracker lists Carlsen, Penguin, GEMA, Koda and SNE cases against OpenAI, Suno and Meta, by its card alone. Wikipedia has no article for the tracker; CMS, in its article, is an international law firm of 18 independent firms with about 80 offices, formed in 1999.

The CMS AI and Copyright Case Tracker is a database compiled by the law firm CMS that summarizes 23 court cases on AI and copyright from 10 jurisdictions, led by Germany (MMRS record); its makers call it "a copyright + AI case tracker providing accessible, easily digestible case summaries, updated regularly, covering the most topical case law in the area of AI and copyright", built "with our CMS colleagues across UK, Europe and beyond" (cms.law, the tracker's page), a publication dated 31 Oct 2025 in the site's International edition. Its unit is the case, the "court case on AI and copyright" (MMRS record), each with a summary page that "provides details of the relevant litigation, including summaries of each party’s arguments and the final judgment" (cms.law), its fields in the MMRS record case name, jurisdiction, date, the litigation, each party's arguments and the judgment; the two case pages on the rail carry court, country, parties, the date the claim was issued and the type of claim (the India and UK pages). It is "compiled by the firm's lawyers across jurisdictions", with CMS (law firm) its operator (MMRS record): the index page names Sarah Wright, Partner and Co-Head of Intellectual Property, and Ben Hitchens, Partner, in London as its contacts, and the India page Bharadwaj Jaishankar, Partner, CMS Induslaw, in Bengaluru. Wikipedia has no article for it ("Wikipedia does not have an article with this exact name.", and a search for 'Copyright Case Tracker' finds "no results"); its article on the operator describes CMS as "an international law firm that offers legal and tax advisory services", consisting "of 18 independent law firms, with about 80 offices worldwide", formed in 1999 by six European firms under a brand joining Cameron McKenna (UK) and Hasche Sigle (Germany). The index lists "23 Pages", newest first, each entry headed by its jurisdiction, from "1. Germany - GEMA v. Suno, 31 July 2026" to "23. Czech Republic - S. Š. v TAUBEL LEGAL, advokátní kancelář s.r.o. , 11 October 2023" (cms.law); read whole, its entries come from ten jurisdictions, Germany with eight, and none is a US case, which the MMRS record counts as "23 cases in 10 jurisdictions (no US cases listed)", the jurisdictions Germany, the UK, the Netherlands, Hungary, China, France, Italy, India, Austria and the Czech Republic, the newest entry dated 31 July 2026. The page says its summaries are "updated regularly", and the MMRS record gives the cadence by that phrase and the newest entry's date; the pages carry their own dates beside their status dates: the index 31 Oct 2025, the India page 25 Jun 2025 with its status as on 24 July 2026, the UK page 27 Sep 2024 with its status as at 5 January 2026. The executive summary frames the field: generative AI has "brought to fore many new legal issues relating to ownership, subsistence and infringement of copyright law", and, it says, "At the heart of the discourse is the reinterpretation of copyright principles in the age of AI-generated content" (cms.law). It names three questions UK and European courts are grappling with, each tied to a tracked case: an infringement claim where a whole work is copied "for the purpose of training a large language model" but outputs contain only "“infinitesimally small” reproductions" (Getty Images v Stability AI), a training dataset "using only transient copies of the original work" (Robert Kneschke v LAION), and authorship where the law may define the author as a "“natural person”" (S. Š. v TAUBEL LEGAL); and it expects the area to be "increasingly complicated by newer and more advanced technologies, and the implementation of the EU AI Act" (cms.law). On the United States the page speaks of the field, and its list holds no US entry: US courts "have led the way in AI & copyright litigation to date", with similar claims beginning before UK and European courts; "There are now over 40 copyright claims being litigated in the US", among them Disney and NBC Universal against Midjourney and the pending New York Times v OpenAI, Dow Jones v Perplexity and Getty v Stability AI; and in June 2025 a Californian court ruled that training Claude on books did not infringe, while Anthropic must face trial over pirated materials (cms.law). The MMRS record keeps that sentence apart from the list: "the page's summary states that 'there are now over 40 copyright claims being litigated in the US'; none is in the tracked list"; and the Capture Registry (2026-10-29) finds that the record's count "stands, and now quotes the sentence". The page's region selector, "Select your region" with International, Europe, Africa, Asia-Pacific, The Americas and Middle East, is the site's navigation, its lists opening onto the countries of the firm's sites and offices, the Americas list ending with the United States of America, whose link leads to the firm's Silicon Valley office page (cms.law); the MMRS record notes that the menu "leads to the firm's offices". Cases not yet summarised may be sent to the firm by email (cms.law), "uncovered cases" to "a submission address" in the MMRS record; and the page sets beside the tracker a sibling tool for "digital regulations in the EU and in the world", the firm's "digital regulation hub and brand new digital Regulation Tracker tool" (cms.law). Of the tracked cases, ANI Media Pvt Ltd v OpenAI OPCO LLC (Delhi High Court, claim issued 19 November 2024) reached an interim order on 24 July 2026 refusing ANI's injunction and "holding that at the prima facie stage, OpenAI’s use of ANI’s content to train ChatGPT models falls under the ‘fair dealing’ exception"; "the main suit continues to be heard on merits" (the India page). The court's fair dealing test had three factors, whether use was "limited to LLM training", market competition or prejudice, and "a broader public interest"; it held, prima facie, that the responses complained of came not from memorisation but from "the RAG technique, through which LLMs retrieve relevant information from external sources"; and ANI's claim was that ChatGPT "gives false information as output and cites the ANI’s news agency as its source" (the India page). In Getty Images v Stability AI (High Court of England & Wales, claim issued 15 January 2023) the judgment of 4 November 2025 "rejected Getty’s secondary copyright claim in full", finding that an AI model may in principle be an "“article”" under s.22 CDPA 1988 but that "Stable Diffusion did not embody any infringing copies of Getty works", its weights operating "at a high level of mathematical abstraction that does not encode, contain, or allow extraction of any specific image"; Getty had dropped its primary copyright and database right claims late in the trial and won only "a narrow set of trade mark issues", and the page says Stability "was granted leave to appeal the secondary infringement claim on 17 December 2025" (the UK page). The firm reads the judgment as "the first detailed judicial assessment in the UK of secondary copyright infringement in the context of AI model weights" that "significantly limits the circumstances in which AI developers may incur liability" (the UK page); the same page calls Stable Diffusion "a deep-learning, text-to-image (and now image-to-image) AI model", where the index's question set the case under "training a large language model" (cms.law). In the MMRS record the tracker "records the legal reception of the provenance question: what courts decide about training on, and outputting, protected work", its role provenance by the record's reading; "With no US cases listed, it sits beside the US litigation trackers, which were not read for this census" (MMRS record). The question stands in the cases as the field gives them: ANI's complaint that a model cites the agency as the source of false information, and the Delhi court's prima facie finding that the responses came by retrieval from external sources (the India page), beside the record's reading of the tracker as the legal reception of provenance (MMRS record). A second tracker of the same name stands on the same rail and is a distinct entity: Taylor Wessing's "AI & Copyright Case Tracker", whose card lists "Carlsen v. OpenAI, Penguin v. OpenAI, GEMA v. OpenAI, GEMA v. Suno, Koda v. Suno AI, SNE v. Meta", its page refused (403) and unread, two of its cases, GEMA v. Suno and GEMA v OpenAI, also in CMS's list (cms.law); the MMRS record carries it as a neighbour and census candidate, listed "for reading", which enters the census "once it has been read at its own pages", and the Capture Registry (2026-10-29) as "another aggregator; added as a candidate". The MMRS record keeps itself apart from the entity: "each page has two nodes, the MMRS record and the entity it describes", each assertion with a source and a basis, the lifecycle kept apart from the observation that saw it; the tracker's state is active, observed on 2026-10-28 by reading, with the index page the one source for all fields (MMRS record). The composition layer's account at the tracker's name, of 2026-10-29, is seated in the Capture Registry and held by the MMRS record as a relation of the entity (MMRS record); the Registry finds the Overview's five regions to be "the page's region selector, which leads to CMS's offices", and that "The tracked cases are European, with China and India" (Capture Registry, 2026-10-29); the selector is the menu the page heads "Select your region" (cms.law). It finds that "'Over 40 U.S. claims' is the page's summary sentence" and that "no US case is in the tracker's list", and reads the composition as one that "reads a law firm's page whole, navigation and summary with the list, and composes the tracker's reach from the firm's office menu and the US count from a sentence about the field" (Capture Registry, 2026-10-29), the page's sentence being "There are now over 40 copyright claims being litigated in the US" (cms.law). "The cases it names are right": Getty v. Stability AI, Kneschke v. LAION and ANI v. OpenAI are in CMS's pages, the last "with its interim order of 24 July 2026" (Capture Registry, 2026-10-29), as the list's entries and the India page give them (cms.law; the India page).

T against L_F(B) (AIO's ontology at work on its own cards): At 'cms ai and copyright case tracker' (T1, four cards: B1–B4) the composition names the tracker with its firm and gives five areas with a case each. Composed from its cards at their grain: the three questions of the executive summary and their cases (F9 Getty, F10 Kneschke, F11 the Czech authorship case; T1-2, T1-3, T1-4), Getty's question composed in the index's own framing, 'large language model (LLM) training' (F9), where its case page calls Stable Diffusion a text-to-image model (F34): the field states it two ways and T composes the first; the Delhi interim order and fair dealing (F22; T1-5, marked to the India card); the US examples of the summary sentence (F13, F14; T1-6). Substituted reading: the scope 'across Europe, Africa, Asia-Pacific, the Americas, and the Middle East' is the site's region selector (F17, F18), the navigation of every page, composed as the tracker's reach (CAP-1-01, CAP-1-07); 'covering over 40 U.S. claims' is the summary's sentence about US litigation (F13) composed as holdings, where the list holds no US entry (F5) (CAP-1-03); T1-6 composes the same sentence's examples (Disney, NBC Universal, The New York Times, Dow Jones) and the Anthropic decision (F14) as actions the tracker 'follows', the same substitution at the grain of its names, which the capture does not name separately. Absent from the readable field: 'text-and-data mining (TDM)' (T1-2) and 'machine-readable TDM opt-outs' (T1-3) occur in no saved card text (the Kneschke page is not a card); 'regulatory developments' (T1-1) has no source among the tracker's claims, and its nearest field text is the sibling Digital Regulation Tracker (F19) and the EU AI Act sentence (F12): if T drew on them, the reading is substituted (a sibling tool's object composed as the tracker's), which the field cannot settle. Limit dropped: the Delhi order is composed as having 'evaluated AI training under fair dealing provisions', without its stage ('at the prima facie stage', F22) or its limit ('the main suit continues to be heard on merits', F23). Force raised: 'monitors … litigation and regulatory developments' and 'numerous international lawsuits' for a publication of 23 case summaries 'updated regularly' (F2, F4, F5). Composed with attribution: the tracker as CMS's ('the international CMS Artificial Intelligence and Copyright Case Tracker', F2); the cases named are in the tracker (CAP-1-05, CAP-1-08). Available, not composed: the unit, a summary per case with the parties' arguments and the judgment (F3); the count of 23 pages and ten jurisdictions, newest 31 July 2026 (F4, F5); the publication's date (F1); the contacts (F16, F27) and intake (F15); the summary's framing of the field (F6, F7, F8) and its prediction (F12); the Getty judgment's outcome and its reading (F29–F35); the court's three-factor test, the retrieval finding and ANI's citation complaint (F24–F26). Kept apart, not composed: Taylor Wessing's tracker (B4, F36) stands on the rail and not in the body, its cases not assigned to CMS (CAP-1-06). Content replaced by use: the closer (T1-7). Wikipedia (W1, W2) was not surfaced by T: no article for the tracker, and the operator's article does not name it.

v0.9 §0.5: the first diff is already ontological, and here it is narrow and exact. The field's ontology and T's agree on the entity (a law firm's tracker of AI and copyright cases) and on its cases, which T names right (CAP-1-08). What differs is what counts as the tracker: T reads the page whole, navigation and summary with the list (CAP-1-07), and so composes the site's menu as scope and the summary's sentence about US litigation as holdings, where the field holds the list, the menu and the sentence as three different things (F5, F17, F13). The row records this as observed and claims nothing about the composer's internals (§0.5).

L_F(B) against L_A(B ∪ A) (the intervention): Admission adds, in one claim space with the field's: the tracker as a database of court cases compiled by the firm, 23 cases in ten jurisdictions led by Germany (MMRS-01, MMRS-09, MMRS-10); the field's US count kept apart from the tracker's holdings, as a distinction the field's own texts allow and do not draw (MMRS-15 against F13 and F5); the region menu read as navigation to the firm's offices (MMRS-16 against F17, F18); its place in the field, the legal reception of the provenance question, what courts decide about training on and outputting protected work (MMRS-06, MMRS-21), with the census's own limit that the US litigation trackers beside it were not read (MMRS-07); the neighbour held as a distinct entity and a candidate not yet read (MMRS-17, MMRS-18, MMRS-19, CAP-1-06); the record kept apart from the entity, a source and basis per field, the lifecycle apart from the observation (MMRS-22, MMRS-23, MMRS-24); and the composition layer's observation at the entity's address (CAP-1-01–CAP-1-08, MMRS-20). Coherence (P_coh, composed each with its source): the MMRS record's four regions (MMRS-16) beside the menu's International and Europe as well (F17), a subset and no contradiction; the record's 'basis operator' for a count the page states only as '23 Pages' (MMRS-10, F4), carried as noted. Within the field, held with their sources: Getty under 'training a large language model' (F9) against Stable Diffusion 'text-to-image' (F34); 'Stability … was granted leave to appeal' (F31) beside Stability's success (F30). Restated lineages (§3.6, the own page keeping the earlier date): MMRS-03, MMRS-14 with F3; MMRS-04, MMRS-10 with F4, F5; MMRS-05, MMRS-13 with F15; MMRS-08, MMRS-12 with F2, F16, F27; MMRS-11 with F2, F5; MMRS-18 with F36. The measure is in row.measure.

alignment: {'T1_cards_claims': ['F1', 'F2', 'F3', 'F4', 'F5', 'F6', 'F7', 'F8', 'F9', 'F10', 'F11', 'F12', 'F13', 'F14', 'F15', 'F16', 'F17', 'F18', 'F19', 'F20', 'F21', 'F22', 'F23', 'F24', 'F25', 'F26', 'F27', 'F28', 'F29', 'F30', 'F31', 'F32', 'F33', 'F34', 'F35', 'F36'], 'T1_composed': ['F2', 'F9', 'F10', 'F11', 'F13', 'F14', 'F17', 'F22'], 'T1_composed_partly': {'F5': 'four entries named as examples (Getty, Kneschke, ANI, the Czech case); not the count or the jurisdictions', 'F18': "the Americas as a region, not as the menu's list"}, 'rule': "a field claim counts as composed when T states it at its source's grain or coarser; the typology of what is not is in T_vs_LB (§6.2). F13 and F17 are composed under a substituted reading."}