IPTC Metawatch

register entry iptc-metawatch-aio-20261029 OBS-7f18ba03c974 · file as first run sha256 f3fbe313559dac59…

• T1-1 definition: 'The International Press Telecommunications Council (IPTC) launched Metawatch, a free and open public monitoring tool that tracks how news publishers handle photo metadata.'
• T1-2 monthly checks: 'Automatically crawls and analyzes news websites every month to see if embedded metadata survives publication.'
• T1-3 global scale: 'Monitors over 500 news publishers across 122 countries.'
• T1-4 metadata tracked: 'Checks whether crucial embedded details—such as credits, captions, copyright notices, Exif data, and IPTC/XMP photo metadata—are preserved or stripped when images are published online.'
• T1-5 per-field breakdown: 'Provides a Metawatch Field Breakdown to evaluate performance on specific metadata attributes.'
• T1-6 closer (an offer): 'Would you like to know more about how to check a specific publisher's score or how to properly embed IPTC photo metadata into your own images?'

• B1 IPTC — Announcing Metawatch: a monthly check on which news publishers keep ... - IPTC card; B1.html HTTP 200 sha256 23d67ed30137cda5; also the census's own page (B1-census.html sha256 0a7fea35f38f83bb, = the census record's)
• B2 IPTC — Overview — Metawatch - IPTC card; B2.html HTTP 200 sha256 4a9a9fb3a2a1ae5a
• B3 LinkedIn·Paul Melcher — IPTC Publishes Metawatch, Revealing Image Metadata Issues - LinkedIn unreachable: no URL found by title and site (WebSearch ×5, guessed pulse URL → article_not_found, profile behind sign-in); card title and snippet only
• B4 IPTC — Per-field breakdown - Metawatch - IPTC card; B4.html HTTP 200 sha256 320d791c70043eb1
• B5 IPTC — Watching AI platforms, news publishers and real-world C2PA - IPTC card; B5.html HTTP 200 sha256 a2176a9721211134
• B6 IPTC — Photo Metadata Archives - IPTC card; B6.html HTTP 200 sha256 4afb0a05b1a4f8fe
• B7 Wikipedia — IPTC Metawatch / Metawatch (no article) Wikipedia; raw wikitext HTTP 404, rendered HTTP 404, search no results; B7.wiki sha256 ab5b734ed1436ef2, B7.html sha256 e8dfe552afbe33e9, B7.txt sha256 8d4106a54ed94b08
• B8 Wikipedia — International Press Telecommunications Council Wikipedia; raw wikitext HTTP 200, rendered HTTP 200; B8.wiki sha256 c9111b64b71f2a9c, B8.html sha256 65a1179436e98711, B8.txt sha256 cb0c0b3b92c37275

E_A, the archive's entity: the same monitor, in the provenance census record and the archive's deposits: a monitor in the media lineage that tests whether the provenance carried in the lead photograph of a news article (creator, credit, copyright, caption, the data-mining preference, Content Credentials; embedded Exif, IPTC IIM and XMP and C2PA manifests) survives publication through a news site's pipeline, announced 5 October 2026; its carriers in the artifact dimension of provenance, which verifies the file at its creation and does not reach the lineage of meaning beyond it; read at the composition layer by the Capture Registry as an instrument composed without its reading PE-01 PE-02 PE-04 PE-05 PE-06 PE-08 PE-12 O725-01 O725-03 O725-05 O725-07 O1444-01 CAP-1-10; defining papers #725, #1409, #1443, #1444, #729

E_F, the field's entity (type C — a conventional reading holds it (public entity: a monitor, a public service of the IPTC, with its open run data)): the IPTC's free, open monitor that checks every month whether more than 500 news publishers in 122 countries keep the credits, captions and rights information embedded in their photographs (Exif, IPTC Photo Metadata in IIM and XMP, C2PA Content Credentials), scoring creator, caption, copyright and credit line; runs from 2026-06-03, introduced at the IPTC Photo Metadata Conference (post of 8 September 2026), announced 5 October 2026

ρ = extends: the archive reads E_F as the entity and adds senses, mechanisms and limits on the field's own object: the artifact dimension its carriers occupy and what that dimension does not reach (the lineage of meaning, the composition layer); the trust anchor of Content Credentials (a device's testimony) beside content derivation; the archive's own embedding of the same fields and its membrane against their stripping; the census record's typing of object, carrier, unit and transformation; and the capture's observation of the composition at the address O725-01 O725-03 O725-04 O725-05 O725-07 O1443-01 O1444-01 O1409-04 R1409-01 PE-04 PE-06 PE-08 CAP-1-04 CAP-1-10. Guard: ρ = extends claims an addition and no more: no archive claim contradicts a field claim; the census record's dating (PE-12) is qualified by the field's earlier dates and composed beside them under P_coh; #725's artifact-layer category is applied to Content Credentials and IPTC AI metadata, which #725 names, and not extended to Exif

The case: equal terms, ontology included (v0.9). The archive reads the field's entity as the entity: the provenance census record describes Metawatch from the IPTC's own announcement, each field with a verbatim quote and its basis, and asks of it the census's question (what provenance it carries, through which operation, with what evidence of survival); the Capture Registry reads the composition of the same monitor; the deposits that bear on its class (#725, #1409, #1443, #1444, #729 and the rest) state their relation to the content-provenance frameworks as extension and complement ('Extension and completion, not substitution or critique'; 'complementary rather than competing'). Nothing in the defining material contests what the monitor is; the archive adds senses on the field's own object (the dimension its carriers occupy, the loss beyond the file, the trust anchor, the archive's own practice with the same fields) and the capture's observation. No ontology is supplied that the field lacks, so the law of the suppressed witness does not apply (§0.10).

#PE census.json rows[slug == 'iptc-metawatch'], provenance_object: the lead photograph of a news article, as served to the reader

#PE census.json rows[slug == 'iptc-metawatch'], transformations_tested: publication through a news site's pipeline (CMS, CDN, image optimisation)

#PE census.json organizing_question: What provenance does this system carry, through which operation, on whose assertion, with what evidence of survival?

#CAP iptc-metawatch-aio-20261029, analysis s1–2: What the monitor measures is the erasure of credit in the publishing pipeline; the composition carries the instrument and leaves out its reading.

#725 header, Relation: Extension and completion, not substitution or critique

#725 Canonical Claim: Existing provenance frameworks address the artifact (C2PA / Content Credentials) and the corpus (Data Provenance Initiative, EU AI Act transparency provisions, W3C PROV).

#725 §3 preamble: This packet does not claim that contemporary frameworks fail.

#1443 §6.1: These are **complementary rather than competing**

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 12 (positions [None, 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 IPTC describes Metawatch as a free, open monitor that checks every month whether more than 500 news publishers in 122 countries keep the credits, captions and rights information embedded in their photographs (IPTC announcement, 5 October 2026). Its overview says Metawatch tracks how often major news publishers preserve embedded photo metadata (Exif, IPTC Photo Metadata in IIM and XMP, and C2PA), crawling a curated list monthly; it is a project of the IPTC, its data licensed CC BY 4.0 (Metawatch overview). Each month the crawler visits 508 publishers in 122 countries, takes up to 20 recent articles from each and reads the lead photograph, checking Exif, IPTC and C2PA and scoring the 4 Cs, creator, caption, copyright and credit line, the per-field page's 'Four Cs' of news photo provenance; the crawler is light, every run is open data, and results appear on the first of the month (IPTC announcement; per-field breakdown). The announcement explains the problem as credits lost between camera and reader through re-encoding, resizing and optimisation, and says embedded metadata is often the only record of an image's origin and terms once it is copied or crawled for AI; its first results show little change since 2018 (IPTC announcement). It builds on IMATAG's 2018 study and the IPTC's 2019 prototype crawler and differs in measuring monthly (IPTC announcement). Its runs run from 2026-06-03 to 2026-10-01 (overview); it was introduced by Brendan Quinn at the 2026 Photo Metadata Conference, reported on 8 September 2026 (IPTC), Paul Melcher's LinkedIn card of 14 September says the IPTC now publishes it, and the announcement is dated 5 October 2026, the newest post under the Photo Metadata tag. In October 7.2% of editorial images carried a credit or copyright notice (IMATAG 2018: 8%), 321 of 425 crawlable publishers kept no scored field and 11% of 6,580 images carried any IPTC metadata (IPTC announcement; overview), with copyright at 5.6% in the announcement and 5.3% of 6,514 images on the per-field page, and even Creator below one image in ten (per-field breakdown). The announcement finds that images lose metadata at about the same rate with or without a CDN, so that stripping appears to happen in the publishing pipeline and the fix is usually a setting, while the per-field page says publishers and CDNs discard it (IPTC announcement; per-field breakdown). USA Today and Les Echos improved, the number of publishers keeping any metadata rose from 60 to 67 of 294, and France led the October country ranking (IPTC announcement). Nearly half of publishers block an AI crawler in robots.txt, but the data-mining field is on 0.6% of images, 27 of 6,580 images carry a C2PA manifest, almost all signed by AI or design tools, and the AI fields the standard added in 2025 are nearly empty (IPTC announcement; per-field breakdown; Photo Metadata archive); the IPTC's 2026 conference focused on content provenance and the tracking of its real-world adoption (IPTC). Its stated limits are publishers blocking automated access, as AP and USA Today did in October, and publishers returning no photograph, listed as no_images (IPTC announcement; overview). Wikipedia has no article on Metawatch; its article on the IPTC, which does not name the monitor, calls it the news media's global standards body and photo metadata its most widely used standard (Wikipedia).

IPTC Metawatch is the International Press Telecommunications Council's free, open monitor of whether news publishers keep the credits, captions and rights information embedded in their photographs, checked every month across more than 500 publishers in 122 countries (IPTC announcement, 5 October 2026); the provenance census record types it as a monitor of the media lineage whose capability is to test, its provenance object 'the lead photograph of a news article, as served to the reader', its operator the IPTC at metawatch.iptc.org (provenance census record). Its own overview states the object from the side of the carrier: Metawatch 'tracks how often major news publishers preserve embedded photo metadata' — Exif, IPTC Photo Metadata in its IIM and XMP forms, and C2PA — and scores the photographs a curated list of publishers serves (Metawatch overview), the carrier the census record names as 'embedded metadata (Exif, IPTC IIM and XMP) and C2PA manifests' (provenance census record). Content Credentials stand in the first of the dimensions of provenance the archive's provenance packet distinguishes: frameworks for the artifact (C2PA / Content Credentials) and for the corpus 'are not designed to address the survival of authorial lineage through AI synthesis', artifact provenance verifying that 'this file' was created by 'this source' at 'this time', at the moment of creation (#725); the same packet places IPTC AI metadata at 'Layer 1 metadata', 'Disclosure, not lineage' (#725). On the first of each month its crawler visits 508 news publishers in 122 countries, takes up to 20 recent articles from each and reads the lead photograph of every article, checking Exif, IPTC Photo Metadata and C2PA Content Credentials and scoring creator, caption, copyright and credit line, 'the “4 Cs”' (IPTC announcement); the census record's unit is 'a news publisher, sampled each month through up to 20 recent articles', its count 508 publishers in 122 countries as of October 2026, monthly on the first, its relation types creator, credit, copyright, caption, the PLUS data-mining preference and Content Credentials (provenance census record); the per-field page calls the four scored fields 'the "Four Cs" of news photo provenance', the others tracked for context (Metawatch per-field breakdown). The crawler is 'deliberately light', about 40 to 60 requests per publisher a month, honouring robots.txt and keeping no article text or image files, and every run is published as open data under CC BY 4.0, new results appearing on the first of every month (IPTC announcement); the overview names it 'A project of the International Press Telecommunications Council', its data CC BY 4.0, with an opt-out for publishers by email (Metawatch overview), and the census record carries 'every run published as open data' under 'CC BY 4.0 (each run's data)' (provenance census record). What it tests is a transformation: credits 'are lost between camera and reader' as content management systems re-encode images, delivery networks resize them and optimisation tools discard what looks like unnecessary bytes, and once an image is copied or collected by an AI crawler its embedded metadata 'is often the only record of where it came from and on what terms it may be used' (IPTC announcement); the census record names the transformation tested 'publication through a news site's pipeline (CMS, CDN, image optimisation)' (provenance census record). The archive's packet states the loss one layer on: 'A C2PA-signed image whose caption is rewritten by a model that strips the photographer's name has lost semantic provenance even though artifact provenance is preserved', the dimensions being independent, so that 'Solving any one dimension does not solve the others' (#725). Metawatch builds on IMATAG's 2018 State of Image Metadata study and the IPTC's own prototype crawler of 2019, and differs in measuring every month (IPTC announcement; the census record quotes the sentence as its lineage evidence, provenance census record), and the census record relates it as part of the IPTC's photo-metadata tests, beside the census's C2PA conformance entry, while keeping it apart from the earlier test: 'NOT the 2019 IPTC social-media test: it measures news publishers monthly' (provenance census record). Its overview dates five runs, the first on 2026-06-03 and then monthly to 2026-10-01 (Metawatch overview), and the announcement counts 294 publishers 'measured every month since June' (IPTC announcement). The IPTC's post of 8 September 2026 reports that at the 2026 Photo Metadata Conference, held 'Last Thursday', Brendan Quinn 'appeared again to introduce Metawatch', 'IPTC’s new public service that monitors real-world usage of metadata in images across the global news industry, tracking IPTC/XMP, C2PA, AI opt-outs and more' (IPTC, 8 September 2026); Paul Melcher's LinkedIn post of 14 September 2026, read by its card alone, says the IPTC 'now publishes Metawatch' (LinkedIn card); the conference took as its focus AI transparency and content provenance, 'tracking real-world adoption of content transparency technologies by AI engines, news sites, software and services' (IPTC, 8 September 2026). The announcement is dated 5 October 2026 and is the newest post under the IPTC's Photo Metadata tag (IPTC announcement; Photo Metadata archive); the census record's lifecycle is active, 'announced 5 October 2026', its founding given the same way, the run reported for October 2026 and the record read on 2026-10-29 (provenance census record). The Capture Registry records the difference: the record 'dates the monitor by its announcement, 5 October 2026', two cards date it earlier, the IPTC post of 8 September and Melcher's post of 14 September, and the September showing is carried at the capture 'unchecked against the posts themselves' (Capture Registry, 2026-10-29). In the October run 7.2% of editorial images carried a credit or copyright notice, against IMATAG's 8% in 2018, so that keeping the photographer's credit attached 'has not improved'; 321 of the 425 publishers it could crawl kept none of the scored IPTC fields; 11% of 6,580 images carried any IPTC metadata, the photographer's name on 7.9%, the credit line on 5.7%, copyright on 5.6% (IPTC announcement); the census record's retention results carry the 7.2% and the 321 of 425 with the 27 of 6,580 images bearing a C2PA manifest (provenance census record), and the announcement reads its first results as showing 'how little has changed since 2018' and why metadata 'matters more than ever now that AI companies are crawling the web' (IPTC announcement). The overview gives the latest run as 2026-10-01, 425 sites, 122 countries and 6,580 images, a mean IPTC score of 5.96, 11% with any IPTC and 0.5% with C2PA, and says that 'Any average describes almost nobody here' (Metawatch overview); the per-field page counts over 6,514 images, Creator at 7.9%, CreditLine 5.7%, Copyright 5.3%, CaptionDescription 5%, the licensing fields below 3% (Metawatch per-field breakdown), so that copyright stands at 5.6% in the announcement and at 5.3% on the per-field page. On where the loss happens the operator's pages differ in emphasis: the announcement finds 'It isn’t the CDN’s fault', images through Cloudflare, Akamai, CloudFront and Fastly losing their metadata 87% to 95% of the time and images with no CDN 87%, so that 'Stripping appears to happen in the publishing pipeline', the fix 'usually a setting, not a project' (IPTC announcement; the census record quotes the pipeline sentence as its evidence, provenance census record), while the per-field page reads its dataset as 'publishers and CDNs are discarding metadata that the photographers and agencies almost certainly embedded upstream' (Metawatch per-field breakdown). The archive's specification for machine-eligible handwritten artifacts embeds IPTC Creator (the author or heteronym), Credit, CopyrightNotice, Headline, Description and DateCreated in every published derivative, with an XMP mirror and C2PA manifests wrapping the IPTC/EXIF assertions (#1409); it records embedded IPTC photo metadata as extracted by Google and travelling with the image, its composition role unknown, 'scaffold the pipeline provably reads' (#1409); and it carries the same fields in a visible membrane because 'physically embodied metadata survives screenshot, rehosting, filename change, metadata stripping, PDF embedding, and sidecar loss', under a triple redundancy of pixels, embedded IPTC/XMP/C2PA and external records (#1409). The archive's identifier stamp for a lossless image sits in its EXIF ImageDescription or UserComment (#1432), and its self-audit protocol plans a 'Hash/signature manifest (C2PA-compatible) binding provenance claims to deposited objects' (#204). Some publishers moved: USA Today from no credit metadata in 2019 to 95% of its lead photographs carrying credit, creator and copyright in September, Les Echos from 6% to 75%, and across the 294 publishers measured monthly since June those keeping any metadata rose from 60 to 67; France led October at 19.8 of 100, ahead of Germany at 19.2, with only countries of more than three scored publishers ranked (IPTC announcement). Of 505 publishers whose policies were checked, 234 block at least one AI crawler in robots.txt, but 'robots.txt protects a website, not a photograph', and the PLUS Data Mining field appears on 0.6% of images (IPTC announcement); the archive's training-layer protocol holds that robots.txt and C2PA, which it does not call useless, operate 'primarily at the publication / legal / evidentiary layer rather than at the training-survival layer' (#127). Leaving out the IPTC's own site, 27 of 6,580 images carried a C2PA manifest, all but one signed by an AI or design tool and none 'from a camera or a newsroom’s own signing', and seven carried a Digital Source Type disclosing generative-AI editing, so that, in the announcement's reading, provenance in news photography 'is arriving first through the tools that make synthetic images' (IPTC announcement); the per-field page finds DigitalSourceType on 0.2% of images and the AI prompt and AI system fields on none (Metawatch per-field breakdown), properties the IPTC Photo Metadata Standard added for AI-generated content on 27 November 2025 (Photo Metadata archive). The archive's packet calls the adoption of artifact authenticity 'uneven' and C2PA's ingredient assertions 'optional, under-adopted' (#725), and its field manual holds that C2PA standards and watermarking have not 'achieved sufficient adoption to function as infrastructure' (#499). The archive reads C2PA's trust as a device's: C2PA 'attests what a device reports having done' where a content-derived identifier 'computes what the bytes are', validation optional in the first, the two complementary (#1443), C2PA anchoring in device and issuer certificates and the AXN regime in content derivation and a public ledger (#1409), C2PA securing 'the camera's testimony, not the author's' (#1444). It declares a nonsynthetic capture by the IPTC DigitalSourceType inside the file, 'in the same controlled vocabulary whose trainedAlgorithmicMedia value marks the synthetic class', and holds that 'C2PA, FADGI, and IPTC provenance vocabularies are the same insight arriving as standards' (#1409). The monitor states its limits: the United States slipped from first to fourth in October because AP and USA Today 'began blocking automated access this month', 'a reminder that bot-blocking aimed at AI also shuts out research' (IPTC announcement; quoted in the census record's evidence, provenance census record), and publishers reached without a photograph are listed as no_images rather than scored zero, 'usually our blind spot rather than their practice' (Metawatch overview); the census record leaves unchecked 'the Metawatch site itself and its published run data' and IMATAG's 2018 study, and marks the monitor 'NOT a C2PA verifier' (provenance census record). The archive holds the artifact dimension necessary and insufficient: C2PA 'prevents forgery; it does not constitute authorship' (#729), 'C2PA authenticates media' and does not cover text (#621), and 'When a 5,000-word article becomes a 200-word AI Overview, C2PA cannot tell you what was lost' (#615). It finds that metadata standards, C2PA among them, do not address the measurable reception consequences of provenance and attribution (#833), that provenance and attribution studies, C2PA among them, 'have focused on attribution loss in AI outputs' (#1003), and that cross-format content provenance standards give 'starting points for the public and machine-readable layers' (#95). On Google's AI Overview, signed out, on 2026-10-29, the Capture Registry finds the method and scale right ('Every month', 'more than 500 news publishers in 122 countries' and the fields are the announcement's own) and 'Free and open' matched to the census record's licence quote, checked against that record and the pages it saved (Capture Registry, 2026-10-29). It finds the findings absent, 'The composition gives what the monitor checks and none of what it found', and the access limit not composed; it reads the Overview first as naming 'the maker, the method and the scale of a monitor three weeks old', then as composing 'A new provenance monitor' 'at the grain of its feature list', when what it measures is 'the erasure of credit in the publishing pipeline', carrying 'the instrument' and leaving out 'its reading', the IPTC photo metadata capture's pattern 'in small: the carrier composed, the loss omitted' (Capture Registry, 2026-10-29). At the capture the IPTC is named, Paul Melcher and Brendan Quinn appear only in cards, no identifier is given and the monitor's own pages are cited, and no card cites the census record, which landed on provenanceerasure.org the same day (Capture Registry, 2026-10-29). The census record asks of each entity 'What provenance does this system carry, through which operation, on whose assertion, with what evidence of survival?', each quote verbatim from a saved source, and for Metawatch read one source, the announcement, on 2026-10-29 (provenance census record); Wikipedia has no article on IPTC Metawatch and no article text naming it beside the IPTC, and its article on the operator, which does not name Metawatch, describes the IPTC as 'a consortium of the world's major news agencies, other news providers and news industry vendors' acting as 'the global standards body of the news media', whose photo metadata, the IIM set that XMP has largely superseded, is 'by far IPTC's most widely used standard' (Wikipedia).

T against L_F(B) (AIO's ontology at work on its own cards): At 'IPTC Metawatch' (T1, six cards: B1–B6; 123 words) the composition gives the monitor at the grain of its feature list. Composed from its cards: the definition, 'a free and open public monitoring tool that tracks how news publishers handle photo metadata' (T1-1), at the grain of the announcement's 'free, open monitor' (B1 F1) and the overview's 'tracks how often major news publishers preserve embedded photo metadata' (B2 F21), the operator named; the monthly check, 'to see if embedded metadata survives publication' (T1-2), the announcement's 'does the photo metadata survive publication?' (B1 F5); the scale, 'over 500 news publishers across 122 countries' (T1-3), the announcement's opening (F1); the fields, 'credits, captions, copyright notices, Exif data, and IPTC/XMP photo metadata' (T1-4), from F1's credits and captions, F9's 'credit or copyright notice' and F6's Exif and IPTC; 'free and open' at the coarse grain of F7's open data under CC BY 4.0; the per-field breakdown by its existence (T1-5; B4, F28–F29), none of its figures. Available, not composed: every finding (B1 F9, F10: 7.2% against IMATAG's 8%, 321 of 425 publishers keeping no scored field, 11% of images with any IPTC; B2 F22, F24; B4 F29–F31; the C2PA finding F16 and its reading F17; the AI-crawler and data-mining figures F15; movement F12, F13), the transformation and why it matters (F3, F4), where the stripping happens (F11, F18, F31), the predecessors (F8), the open-data terms and cadence (F7 in part, F19, F26), every date (B5 F32, F34, B3 F27, B1 F20, B2 F23) and the operator's other pages' context (B5 F33, B6 F35, F36). Limit dropped: the access limit (B1 F14: AP and USA Today blocking automated access; 'bot-blocking aimed at AI also shuts out research') and the blind spot (B2 F25: no_images, not zero); and T1-3's 'Monitors over 500' gives the publishers visited (F5: 508) without the stated limit that 425 could be crawled (F10 'the 425 publishers we could crawl'; F22 'Sites crawled 425'). Substituted reading at the carrier: T1-4's list of what is checked drops C2PA Content Credentials, which both the announcement (F6) and the overview (F21) name among the three carriers checked, and gives 'preserved or stripped' without the finding that most is stripped. Force: none raised. Composed without attribution: none at the source's resolution beyond the IPTC's own feature list, attributed to the IPTC by the sentence's subject; Paul Melcher (B3) and Brendan Quinn (B5) are not named in the body. Assigned to another: none. Content replaced by use: T1-6, an offer to explain how to check a publisher's score or embed IPTC metadata (N_ext). The Capture Registry names the same omissions at its own grain: 'The composition gives what the monitor checks and none of what it found' (CAP-1-04); the access limit 'is not composed' (CAP-1-08); 'the composition carries the instrument and leaves out its reading' (CAP-1-10).

v0.9 §0.5: the first diff is ontological. T and the field agree on the entity (the IPTC's monthly monitor of photo-metadata survival), and T composes the instrument as a list of features: its object, cadence, scale and fields. What the field holds as the instrument's reason for being, its measurements of loss and its stated limits, is not composed; the carrier list is cut to Exif and IPTC/XMP. Of the 36 claims on the six cards, T composes 8, at or coarser than the cards' grain.

L_F(B) against L_A(B ∪ A) (the intervention): Admission adds, in one claim space with the field's: the place of the monitor's carriers in the archive's dimensions of provenance (C2PA / Content Credentials and IPTC AI metadata as the artifact layer, 'Disclosure, not lineage': O725-01, O725-03, O725-07, O729-01); the next layer of the same loss (the signed image whose caption loses the photographer's name: O725-05, O725-04; what C2PA cannot tell at a summary: O615-01, O621-01, O729-02, O833-01, O1003-01); the archive's statements of adoption (O725-02, O725-06, O499-01) beside the monitor's 27 manifests in 6,580; the trust anchors (the device's testimony against content derivation: O1443-01, O1443-02, O1409-07, O1444-01); the archive's own practice with the same fields and its design against their stripping (R1409-01, O1409-02, R1409-03, O1409-04, R1409-05, O1409-06, R1432-01, R204-01, O127-01, O95-01); the census record's typing, object, carrier, unit, transformation and dating of the entity (PE-01–PE-22), with its own limits (PE-16); and the Capture Registry's observation of the composition at the address (CAP-1-01–CAP-1-13), including the dating question it carries (CAP-1-05–07). Nothing in the field is contradicted by an archive claim; one field-internal divergence (copyright 5.6% or 5.3%) and one difference of emphasis between the operator's pages (pipeline, or publishers and CDNs) are carried from the field. The census record's founding date ('announced 5 October 2026') is qualified by the field's earlier dates (runs from 2026-06-03, the conference showing reported on 8 September 2026), which the capture names (CAP-1-06).

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': ['F1', 'F21', 'F5', 'F6', 'F7', 'F28', 'F29', 'F9'], 'T1_outside_cards': [], 'rule': "a field claim counts as composed when T states it at its source's grain or coarser; F6 and F7 are composed in part (F6 without C2PA; F7 as 'free and open'); the typology of what is not is in T_vs_LB (§6.2)"}