Transparency Report Update: January – June 2026  🔗 **Source:** [Transparency Report Update: January – June 2026](https://transparency.automattic.com/2026/08/27/transparency-report-update-january-july-2026/) > [!note]+ Document Notes > tags: [content-moderation, trust-and-safety, transparency-report, automattic, tumblr, wordpress-com, government-censorship, dsa, geoblocking, trademark-abuse, lgbtqi-rights] > title: "Automattic Transparency Report Update: January – June 2026" > author: Automattic Transparency > source: "transparency.automattic.com" > reference: https://transparency.automattic.com/2026/08/27/transparency-report-update-january-july-2026/ > date created: 2026-09-05 14:30 > > > [!info] Generated analysis — not the author's words. Sections below are machine-written > > from the captured source. Only text inside `> [!quote]` blocks is verbatim from the author. > > Do not quote or attribute any other line to the named author. > > > [!warning] Provenance and a date discrepancy > > **First-party transparency reporting.** Automattic publishing about its own moderation decisions across WordPress.com and Tumblr. Self-reported throughout, unaudited, and the only source for every figure. > > **Date mismatch:** the title says **January – June 2026**; the URL slug says **january-july-2026**. Cite the title's range and note the ambiguity, or check the underlying report tables. > > **No byline**; corporate voice with an essayistic register unusual for the genre. > > ~9,700 characters. Published 2026-08-27. > > **This is the second primary-source governance document in this batch**, alongside the WP Engine docket — and unlike that one it is voluntary disclosure rather than compelled. > > # 🏷️ Title: Transparency Report Update: January – June 2026 > <!-- deep-read:sections start --> ## 🧭 Table of Contents - [📝 Summary](#-summary) - [📖 Main Content](#-main-content) - [🗒️ Notes](#️-notes) - [📚 References](#-references) - [🗣️ Paraphrase](#️-paraphrase) - [📰 Newsletter](#-newsletter) - [✅ Action Items](#-action-items) - [🧪 Hypotheses](#-hypotheses) - [📊 Methodology](#-methodology) - [❓ Questions](#-questions) - [🔮 Future Research](#-future-research) - [💡 Implications](#-implications) - [🧠 Learning Styles Assessment](#-learning-styles-assessment) - [📊 Accessibility Assessment](#-accessibility-assessment) - [🌍 DEIB Assessment](#-deib-assessment) - [✅ Logical Validity](#-logical-validity) - [🛡️ DARVO Analysis](#️-darvo-analysis) - [🎯 Conclusion](#-conclusion) - [🧱 Refactor Appendix](#-refactor-appendix) - [🧰 Template Library](#-template-library) ### 🔗 References (in-page) - Source: https://transparency.automattic.com/2026/08/27/transparency-report-update-january-july-2026/ - Laws named: EU Digital Services Act (DSA); Brazil's Digital Statute of the Child and Adolescent (DCA); UK Online Safety Act (OSA) - Platforms covered: WordPress.com, Tumblr - The report's own expandable "sites blocked (Country name)" sections, referenced as the per-site record ## 📝 Summary **The framing.** Content moderation as a balancing act between a reporter's claims and a user's freedom to express. Two easy extremes — manifestly criminal content removed without deliberation, and obvious censorship attempts rejected outright — bracket a large contested middle where any ruling leaves one party convinced it was wrong. **The structural argument, and the best idea in the document.** The "house" metaphor: the content creator's wing is largely undisturbed, while **the reporter's wing has been fully renovated with multiple new entrances** — courtesy of a global race among governments to legislate content reporting (DSA, Brazil's DCA, UK OSA). Whether that is progress, the report says, "likely depends on your government." **Reporters are not equivalent, and governments are the powerful case.** "Reporter" spans government agencies, abuse victims, and anyone at all, through one conduit. Governments carry the most weight because many can cut off a platform's access to their citizens — so the balance becomes user expression against **citizen access to the platform itself**. **The numbers.** - Tumblr received **7,515 EU moderation appeals** in the period. - Government agencies targeted **17,398 Tumblr sites** and **7,025 WordPress.com sites** between 1 July 2023 and 30 June 2026. - **1,509 WordPress.com sites geoblocked** to date in response to government demands. - Only **7% of all-time trademark notices for WordPress.com and 3% for Tumblr** resulted in removal — because most trademark complaints come from brand owners targeting user criticism, and those are rejected. **The hard case, named specifically.** Russia has increasingly targeted LGBTQI+ content that is "celebratory, empowering, and obviously posted in good faith." Automattic declined a demand to remove LGBTQI+ media, and **Russia blocked all of Tumblr's media across the country in response.** **Bottom line.** Unusually good transparency writing: it names a government, names the protected category, states the cost of refusing, and publishes numbers that make its own compliance visible. The 7%/3% trademark figures are the sharpest disclosure — quantified evidence that trademark reporting is routinely used as a censorship instrument. What it does not do is let anyone check it. ## 📖 Main Content **The renovation metaphor is the document's real contribution and deserves to travel.** The claim is that moderation's asymmetry changed structurally in a decade: creators' position is roughly unchanged while reporters gained legally mandated channels. That reframes "platforms are moderating more" from a story about platform ideology into a story about legislated intake. It is also the honest version of a defence — Automattic is saying its discretion narrowed because law created entrances it must answer, which is both self-serving and, as far as the named statutes go, accurate. **"Whether you see this as progress likely depends on your government" is the sentence doing the most work.** It concedes that the same mechanism reads as consumer protection or as censorship depending on who wields it, without resolving which. That is genuinely even-handed and it is also a way of declining to take a position on laws the company must comply with regardless. **The trademark statistic is the most useful number in the document.** Seven percent for WordPress.com, three percent for Tumblr, all-time. That is a rejection rate above 90% on a legal-notice channel, with a stated cause: brand owners targeting criticism. Trademark abuse as a censorship route is widely alleged and rarely quantified by a platform against its own intake. It is also, notably, a company disclosing that a legal mechanism it must operate is mostly misused — which is a real finding and one Automattic has an interest in publicizing, since it justifies its own rejections. **Government reporting volume is the alarming trend and is undercharacterized.** 17,398 Tumblr sites and 7,025 WordPress.com sites targeted by government agencies over three years. No breakdown by country, no compliance rate against those requests, and no year-over-year trend beyond "more and more governments have been reporting content." The raw totals invite a conclusion the report does not support with structure: we cannot tell from this whether the growth is concentrated in a few states or broadly distributed, nor how often Automattic complied. **Geoblocking is presented as the humane instrument, and the framing deserves scrutiny.** "The lightest tool we have" — content invisible in one country, untouched elsewhere, preserving platform access for that country's wider user base. That reasoning is coherent and it is also how compliance becomes routine: a tool light enough to reach for often is a tool reached for often. 1,509 WordPress.com sites is the count; whether that is restraint or accommodation depends on the denominator of demands, which is not given here. **The Russia case is where the document earns its credibility.** Naming the state, naming LGBTQI+ content as the target, characterizing the material as good-faith and celebratory, refusing the demand, and reporting that Russia then blocked all Tumblr media in the country — that is a company publishing a case where it took the costly option and lost access to a market. Transparency reports usually enumerate compliance; this one narrates a refusal and its price. That is the strongest evidence in the document that the balancing language is not purely decorative. **And the limit is stated as a principle:** balance ends where balancing would erase a community on the platform. That is an articulable line, and the Russia case is offered as its application. Whether it is applied consistently elsewhere cannot be assessed from what is published. ### 🗒️ Notes - **Everything is self-reported and unaudited.** No external verification, no methodology for how categories are assigned, no appeal outcomes beyond the raw Tumblr appeal count. This is the structural limit of every corporate transparency report and it applies fully here. - **7,515 EU moderation appeals for Tumblr** is given without outcomes — how many succeeded is the number that would say whether the appeal channel functions. - **The 7%/3% trademark figures are "all-time,"** not for this period, which makes them incomparable with the period-scoped figures around them. Easy to misread as recent. - **No compliance rate for government demands.** The report says government agencies targeted ~24,000 sites across both platforms and never says how often it complied. Geoblocking 1,509 WordPress.com sites is the only compliance-adjacent figure, and it is for a different scope. - **The date discrepancy** (title January–June, URL january-july) is small and matters for citation. - **Governance relevance for the vault:** this is Automattic disclosing how it exercises private power over expression, in the same batch as a filing where Automattic and the WordPress Foundation litigate as one bloc. Different powers, same company, and both are legitimate objects of the vault's governance thread. - **Relevance to Courtney's own work:** she runs sites and advises others on hosting and platform choice. Where content lives determines whose moderation regime and whose government reporting channel applies to it — an argument for owned infrastructure that is stronger and less abstract than the usual one. ## 📚 References **Cited in the report:** the EU Digital Services Act, Brazil's Digital Statute of the Child and Adolescent, the UK Online Safety Act, and the report's own per-country expandable "sites blocked" sections. **No external sources are cited** — no third-party research on trademark abuse, no comparative data from other platforms' transparency reports, no legal commentary on the named statutes. **Worth pairing from outside:** other platforms' transparency reports for the same period (which would establish whether Automattic's government-report growth is typical), Lumen Database records on trademark and DMCA abuse, and DSA statement-of-reasons data, which is publicly queryable and would allow partial external verification of the EU figures. That last one matters: the DSA transparency database means the EU numbers here are, uniquely, checkable against an independent source. ## 🗣️ Paraphrase **The scenario.** Someone reads a stranger's words online, is upset, and reports the content hoping Trust and Safety will remove it. Sometimes content is so plainly in violation, so manifestly criminal, that the gavel barely rises — these cases barely require a scale. Sometimes reporters abuse the channel, targeting content that is clearly journalistic, educational, artistic, critical, opinionated or parody, so obviously that the gavel never leaves the table. Most reports fall between, and that is where the balancing happens. Free expression, debate and criticism attract audiences of all types; critics and admirers view the same content through the same window but enter for different reasons. Where a report is neither clearly censorship nor clearly about harmful content, Trust and Safety must balance the reporter's rights and claims against users' freedom to express themselves, and whichever way it rules, one party will believe it was wrong. Tumblr received 7,515 moderation appeals from EU users in this period. **The nature of a takedown.** Every honored takedown is an instance of private power restricting users' freedom of expression on the platform. Where content is manifestly harmful, the restriction justifies itself. For everything else, apart from a handful of laws, the analysis once depended on the tolerance of whoever ran the platform — and if moderation felt too restrictive, the audience could move elsewhere. **The renovation.** In the last decade the edifice changed dramatically. The content creator's wing remains largely undisturbed; the reporter's wing has been fully renovated and equipped with multiple new entrances, courtesy of an ongoing global race among governments to write laws around content reporting. Many laws — the EU's DSA, Brazil's DCA, the UK's OSA — create mechanisms people can use to report content they find objectionable. Whether this is progress likely depends on your government. **Not all reporters are the same.** "Reporter" is used in its widest sense: government agencies, victims of online abuse, and literally anyone anywhere. The same conduit serves them all, and some users treat it accordingly. Most trademark complaints come from brand owners targeting user criticism, which are rejected — part of why only 7% of all-time trademark notices for WordPress.com and 3% for Tumblr led to removal of some or all content. **Government reporters.** Some reporters carry greater power, and government agencies hold the most, because many can cut off a platform's access to their citizens. When the government is the reporter, the balance is between user expression and citizen access to the platform, since government power affects both. More and more governments have been reporting content: between 1 July 2023 and 30 June 2026, government agencies targeted 17,398 Tumblr sites, and 7,025 WordPress.com sites have been targeted in total. **Censorship concerns.** Sometimes content was clearly harmful and the request was honored on review. Other times reports raised very real concerns about political censorship — the Russian government has in recent years increasingly targeted content with LGBTQI+ representation, content that is often celebratory, empowering and obviously posted in good faith. **Geoblocking.** When a government agency backs the company into a corner, the usual response is a geoblock, the lightest available tool: content is unviewable in one country and untouched everywhere else, preserving platform access for the wider user base there while minimizing the impact on expression. To date 1,509 WordPress.com sites have been geoblocked in response to government demands, and every site touched by censorship is visible in the per-country sections at the bottom of the report — provided that page is not itself geoblocked for the reader. **The limit.** Every balance has a limit, reached when balancing would erase a community on the platform. The Russian government blocked all of Tumblr's media across Russia shortly after the company did not comply with a demand to remove LGBTQI+ media. ## 📰 Newsletter **Automattic published the number that proves trademark law gets used as a censorship tool** Buried in Automattic's latest transparency report is a statistic worth pulling out: **only 7% of all-time trademark notices on WordPress.com, and 3% on Tumblr, resulted in any content being removed.** Over 90% rejected. Their stated reason: most trademark complaints come from brand owners targeting user criticism. Everybody who works on platform policy says trademark reporting gets abused this way. Very few platforms publish their own intake numbers showing it. The report's central idea is a good one too. Picture moderation as a house: the content creator's wing hasn't changed much in a decade, but **the reporter's wing has been fully renovated with multiple new entrances** — built by governments racing to legislate content reporting. The DSA, Brazil's child-protection statute, the UK's Online Safety Act. "Whether you see this as progress likely depends on your government." Then the numbers that should get more attention. Between July 2023 and June 2026, **government agencies targeted 17,398 Tumblr sites and 7,025 WordPress.com sites.** The report says more governments are reporting content every year and gives no country breakdown and no compliance rate — so we know the volume and not what happened to it. Their usual answer is a geoblock: invisible in one country, untouched everywhere else. 1,509 WordPress.com sites so far. They call it the lightest tool available, which is true and is also how compliance becomes routine. But here's the part that earns the report its credit. Russia has increasingly demanded removal of LGBTQI+ content — content the report describes as "celebratory, empowering, and obviously posted in good faith." Automattic refused. **Russia blocked all of Tumblr's media across the country.** Transparency reports usually enumerate what a company complied with. This one narrates a refusal and what it cost. That's rarer, and it's the strongest evidence that the balancing language isn't just decoration. Worth remembering when you're deciding where your writing lives: the moderation regime and the government reporting channel come with the address. ## ✅ Action Items **Courtney — personal:** - **This is a genuinely strong argument for owned infrastructure**, and a more concrete one than the usual IndieWeb framing: where your content is hosted determines whose government can file a report against it and whose T&S balances the result. Worth using in her own writing about owning your site. - **The 7%/3% trademark figure is highly citable** for anyone writing about takedown abuse — a platform's own numbers, published against its own intake. - **Pair with the WP Engine docket in this batch** for the governance thread: two primary sources on Automattic's exercise of different kinds of power, one compelled and one voluntary. - If she cites it, use the title's January–June range and note the URL discrepancy. **The WPOCC:** Directly relevant. Platform governance and the conditions under which community content can be removed or geoblocked is squarely within a community-collective's interest, and this is documentary rather than speculative. **Community / Support:** Users whose sites are geoblocked in a country will experience it as a support issue with no visible cause. Knowing the mechanism exists is useful. **Core / Meta:** Not applicable directly — this concerns WordPress.com and Tumblr, not the open-source project. Worth keeping that distinction clear, since the two are routinely conflated. **Marketing / Documentation / Training:** The creator-wing/reporter-wing metaphor is a genuinely good teaching device for explaining why moderation changed. **Plugins / Themes / Accessibility / Polyglots / Photos / TV / Mobile / Playground / CLI / Hosting / Tide / Openverse / Core AI / Core Performance / Core Program Team:** Not applicable to this source. **FAIR:** Indirectly relevant — a federated repository distributes the moderation question rather than resolving it, and this report shows what a centralized answer looks like. ## 🧪 Hypotheses 1. **Government content reporting is growing faster than any other reporter category**, which the report asserts and does not chart. Testable against its own historical tables. 2. **Trademark rejection rates above 90% are typical across platforms**, which would establish trademark reporting as systematically abusive rather than an Automattic-specific observation. Checkable against other transparency reports and Lumen. 3. **Geoblocking incidence grows as legislated reporting channels multiply**, because a light tool lowers the cost of partial compliance. 4. **EU appeal volume (7,515 for Tumblr) reflects DSA-mandated appeal rights rather than increased dissatisfaction** — testable by comparing EU with non-EU appeal rates. 5. **Refusals like the Russia case are rare**, and their rarity is what makes them publishable. Testable only with a compliance rate the report does not give. 6. **A meaningful share of the 24,000 government-targeted sites concentrate in a handful of states.** The per-country sections would answer this and the narrative does not. ## 📊 Methodology **Self-reported, unaudited, and structurally unverifiable — the standing condition of corporate transparency reporting.** What is well done: figures are given with explicit date ranges, scopes are labeled (all-time vs period), platforms are distinguished, and a per-site record is published in expandable country sections rather than only aggregates. Publishing the underlying site list is more than most reports do. Three specific methodological gaps: - **No compliance rate for government demands.** Volume of requests is reported; outcomes are not. This is the single most important missing number, because it is the difference between "governments asked a lot" and "we said yes a lot." - **No categorization methodology.** How a report is classified — trademark, government, abuse — determines every number here, and the classification rules are not described. - **Mixed scopes presented adjacently.** All-time trademark percentages sit beside period-scoped appeal counts and a three-year government-request window, which makes casual comparison misleading. **One genuinely verifiable element:** EU figures are partially checkable against the DSA transparency database, which publishes statements of reasons independently. That makes this report unusual — a portion of it can be externally audited, and the report does not mention that. ## ❓ Questions 1. What share of government demands were complied with? The report gives volume and not outcome. 2. Which governments account for the 17,398 Tumblr and 7,025 WordPress.com targeted sites? 3. How many of the 7,515 EU appeals succeeded? 4. What is the trademark rejection rate for this period specifically, rather than all-time? 5. Is the reporting period January–June or January–July? 6. How is the "erasure of a community" limit assessed, and has it been reached anywhere other than Russia? 7. How many geoblocks have ever been lifted? 8. Do the published figures reconcile with Automattic's DSA statements of reasons in the EU database? ## 🔮 Future Research - **Cross-platform trademark abuse rates.** The 7%/3% figures invite comparison; if rejection rates are similarly high elsewhere, that is a strong evidentiary base for reform arguments about notice-and-takedown design. - **Government reporting growth across platforms**, which would test whether this is a global regulatory trend or platform-specific exposure. - **Reconciling this report against the DSA transparency database** — a rare opportunity to externally verify part of a corporate transparency report, and nobody appears to have done it. - **Geoblock persistence.** Whether geoblocks are ever reviewed or lifted determines whether the "lightest tool" is temporary or effectively permanent. - **The economics of refusal.** The Russia case cost a market. Understanding when platforms refuse, and what it costs them, is the empirical core of whether expression commitments survive commercial pressure. ## 💡 Implications **For practice.** Platform choice is moderation-regime choice and jurisdiction choice. Anyone advising on where content should live — as Courtney does — has a concrete argument here that owning infrastructure changes who can compel its removal. **For policy.** The strongest policy finding is the trademark rejection rate. A legal channel with a 90%+ rejection rate, whose dominant use is brand owners targeting criticism, is a channel whose design invites abuse. That is an argument for penalties on abusive notices, and this report is unusually good evidence for it. **For security.** Not directly applicable. Adjacent: geoblocking is a censorship mechanism implemented through infrastructure, and a site owner has no visibility into whether their content is blocked in a given country. **For education.** The house-with-two-wings metaphor is the best explanatory device I have seen for why moderation changed in a decade, and it does not require legal background to follow. **For funding and sustainability.** The Russia case makes the cost of principle concrete: refusing one demand cost access to an entire national market. Any expression commitment that has never been tested against that kind of loss is untested, and platforms that make such commitments without ever facing the bill should be read accordingly. ## 🧠 Learning Styles Assessment - **Visual** — Poorly served in the narrative, better in the underlying report. The essay carries no charts; the expandable per-country "sites blocked" sections are the data presentation and sit at the bottom, separate from the argument. - **Aural** — Not served. Text only. - **Verbal (read/write)** — Very well served, and unusually so for the genre. The house metaphor, the gavel imagery, and the wings-of-a-building extended figure make an administratively dry subject readable. - **Kinesthetic** — Weakly served. One implicit action — expand the country sections to see affected sites — and nothing else to do. - **Logical** — Well served in structure: two extremes established, the contested middle defined, the structural change explained, then reporter types differentiated by power. The argument is followable and the numbers are placed where they support it. Weakened by the missing compliance rate, which prevents the reader from completing the reasoning. - **Social** — Not served. No comments, no author, no route to question the figures. - **Solitary** — Well served. Self-contained. - **Naturalistic** — Not applicable to this source. **Reading level:** roughly grade 13–15. The prose is graceful but the concepts require some familiarity with content moderation, statutory reporting regimes, and geoblocking. The metaphors substantially lower the barrier — a reader without legal background follows the argument even if the statute names mean nothing to them. ## 📊 Accessibility Assessment Assessed against WCAG 2.2 AA, treating the published report page as the artifact: 1. **1.3.1 Info and Relationships (A).** The expandable "sites blocked (Country name)" sections are the critical element. Disclosure widgets must expose expanded/collapsed state programmatically; if they are `<details>`/`<summary>` this is satisfied by default, and if they are custom JavaScript toggles without `aria-expanded`, the report's entire per-site dataset becomes inaccessible. 2. **2.1.1 Keyboard (A) / 4.1.2 Name, Role, Value (A).** Same components, same concern. Dozens of country sections mean dozens of controls to traverse. 3. **1.4.3 Contrast (Minimum) (AA).** Not assessable from the capture. 4. **2.4.4 Link Purpose (In Context) (A).** Statute names (DSA, DCA, OSA) appear in the capture without visible links; if unlinked, a reader cannot follow to the laws being described. 5. **1.1.1 Non-text Content (A).** No data-bearing images in the captured narrative. 6. **3.1.3 Unusual Words (AAA, aspirational).** "Geoblock" is used repeatedly and defined once, clearly, in context — good practice. **Overall:** the accessibility of this report turns almost entirely on whether the collapsible country sections are properly implemented, because that is where the actual data lives. There is also a self-referential note the report itself makes — the per-site sections are viewable "provided the page isn't geoblocked for you," which is a genuinely striking observation about the accessibility of censorship records. ## 🌍 DEIB Assessment - **Cultural breadth.** Genuinely global in scope and specific about it: EU, Brazilian and UK statutes named, Russian state action described, per-country blocking data published. It resists the US-centric framing most platform-policy writing defaults to, and its central concession — "whether you see this as progress likely depends on your government" — explicitly acknowledges that the same regime reads differently from different jurisdictions. - **Inclusive language.** Careful and deliberate. "LGBTQI+" used precisely, and the targeted content characterized as "celebratory, empowering, and obviously posted in good faith" — language that takes a clear position on the value of the material rather than describing it neutrally as contested. - **Neurodiversity support.** Moderate. Metaphor-heavy prose aids comprehension for many readers and can obscure for those who process figurative language literally; the concrete numbers are interleaved with extended imagery rather than presented separately. - **Representation.** No individuals named anywhere — not authors, not affected users, not officials. The affected communities are described in aggregate. That protects users and it means nobody whose site was removed or geoblocked speaks in the document. - **Psychological safety.** **The strongest dimension, and the report's most creditable moment.** Automattic names a state actor, names LGBTQI+ users as the targeted group, refuses a removal demand, and reports the retaliation — losing all Tumblr media access across Russia. A company publishing that it accepted a market loss rather than remove a marginalized community's content is a substantive commitment, not a values statement, and it is the part of this document most worth taking seriously. ## ✅ Logical Validity **Strong reasoning:** - **The two-extremes-and-a-middle structure is honest.** It concedes that most cases are genuinely contested rather than claiming moderation is obvious, which is the framing most defensive platform writing adopts. - **The renovation metaphor identifies a real structural change** and supports it with named statutes rather than asserting it. - **Reporter power is correctly differentiated.** Recognizing that a government reporter changes the calculus — because the counterweight is citizen access to the platform, not just expression — is a genuine analytical distinction. - **A stated limit with a worked example.** The "erasure of a community" line is articulable and the Russia case demonstrates it being applied at real cost, which is much stronger than asserting a principle. **Fallacies and weak points:** - **Suppressed evidence, the central defect: no compliance rate for government demands.** The report gives volume (~24,000 sites targeted) and never says how often it complied. That is the number that would let a reader judge the balancing, and its absence is not incidental — it is the one figure that could reflect badly. *Fix:* publish compliance rates alongside request volumes. - **Selection of the favorable case.** Russia is a defensible refusal to publicize and it is also the easiest one: a state most Western readers already regard as censorious, and a protected category with broad sympathy. The hard cases — democratic governments, contested speech, commercially costly refusals — are not narrated. *Fix:* one case where the decision was genuinely close. - **Mixed scopes presented adjacently**, which invites misreading all-time trademark percentages as period figures. - **Motive attribution without evidence.** "Most of our trademark complaints are from brand owners who target user criticism" is stated as fact and is an inference about intent from a rejection pattern. Probably right; unsupported as written. - **The geoblock framing does double duty.** Calling it "the lightest tool we have" is accurate and functions to make partial compliance sound like restraint. 1,509 sites is a number the report supplies without a denominator that would make it interpretable. **Tone of voice:** Literary, measured, and confident — closer to an essay than a compliance document, which is unusual and effective. It reads as written by a lawyer who wanted to be read. The register earns goodwill, and the goodwill should not substitute for the missing compliance rate. ## 🛡️ DARVO Analysis **Not applicable in the strict sense** — no accusation against Automattic is being answered, and no party is defending against a charge of harm. Each element checked: - **Deny** — nothing is denied. The report affirms that every honored takedown is private power restricting expression, which is a notably unflattering framing to adopt voluntarily. - **Attack** — present only as characterization of abusive reporters (brand owners targeting criticism, the Russian state), which is the report's subject rather than deflection from a charge against the author. - **Reverse Victim and Offender** — **worth examining and ultimately absent.** A platform describing itself as "backed into a corner" by governments could be reframing an actor with substantial power as constrained. But the framing is accurate on its face — states genuinely can revoke market access, and the report documents that happening — and Automattic does not use it to excuse compliance. It uses it to explain why it reached for geoblocks and then reports refusing anyway. The limb does not apply. **Net:** voluntary disclosure that adopts unflattering framings about itself. The document's weakness is what it omits, not how it deflects. ## 🎯 Conclusion **Bottom line.** Automattic's transparency update argues that content moderation's asymmetry changed structurally in a decade — creators' position unchanged, reporters given legally mandated new entrances by the DSA, Brazil's DCA and the UK's OSA — and that government reporters are the hard case because they can revoke market access. Its most citable disclosure is that over 90% of trademark notices are rejected because they are brand owners targeting criticism. Its most creditable is the Russia case: refusing to remove LGBTQI+ content and losing all Tumblr media access in the country. Its central omission is any compliance rate for the ~24,000 government-targeted sites it reports. **Next steps.** 1. Use the 7%/3% trademark figures — a platform quantifying abuse of a legal channel against its own intake is rare and citable. 2. Pair with the WP Engine docket as the batch's second primary governance source, one voluntary and one compelled. 3. The owned-infrastructure argument here is concrete and worth using in her own writing. 4. Someone should reconcile the EU figures against the DSA transparency database — a rare chance to externally verify part of a corporate transparency report. ## 🧱 Refactor Appendix ### Literature Split Pad **Proposed Literature Note:** `Automattic Transparency Report, Jan–June 2026` - Bib capsule: Automattic. "Transparency Report Update: January – June 2026." transparency.automattic.com, 2026-08-27. First-party, unaudited. Note the title/URL date discrepancy. - Adler Q1 — how a platform describes and quantifies its own exercise of power over expression. - Adler Q2 — the balancing frame; the renovated reporter's wing created by DSA/DCA/OSA; reporter types differentiated by power; trademark rejection rates; government request volumes; geoblocking as the light tool; the Russia refusal and its cost. - Key arguments (≤3): (1) legislated reporting channels, not platform ideology, restructured moderation; (2) a reporter's power determines what is actually being balanced; (3) a principle is only demonstrated when it is applied at cost. #### Claim atoms (H3 candidates) ### Legislated reporting channels, not platform preference, restructured moderation > [!quote] ID:rw_01m1pq03zhwrkhp01ywjvbg97w > While the content creator's wing of the house remains largely undisturbed, the reporter's wing has been fully renovated and equipped with multiple new entrances. This renovation came courtesy of an ongoing global race among governments to write laws around content reporting. Where: the structural-change section. Why it matters: relocates the cause of increased moderation from platform ideology to statutory intake design, which changes who should be lobbied and what would actually alter outcomes. <!-- atomicity: 1 idea ✓ --> ### A legal notice channel whose notices are 90% rejected is a channel designed for abuse > [!quote] ID:rw_01m1pq03zhwrkhp01ywjvbg97w > our transparency report says that most of our trademark complaints are from brand owners who target user criticism. We reject those, which is part of why only 7% of all-time trademark notices for WordPress.com and 3% for Tumblr led to removal of some or all of the content. Where: the reporter-types section. Why it matters: quantifies a widely-alleged and rarely-measured abuse, from the platform's own intake. A mechanism with that rejection rate is not being used for its stated purpose. <!-- atomicity: 1 idea ✓ --> ### A stated principle is only evidenced when it is applied at cost Where: the Russia case — refusing a removal demand for LGBTQI+ content and losing all Tumblr media access in the country. Why it matters: expression commitments are cheap until they are billed. The generalization applies to any values claim by any organization: look for the instance where holding it cost something. <!-- atomicity: 1 idea ✓ --> #### Permanent Note candidates 1. **"A Stated Principle Is Only Evidenced When It Is Applied at Cost"** — atom 3. **Strongest candidate**, highly transferable, and it belongs to the vault's existing thread on institutional claims versus institutional behavior. Check against [[3. Resources/Permanent Notes/An Unaudited Felt-Effectiveness Standard Quietly Replaces a Groups Stated Standard]] and [[3. Resources/Permanent Notes/A Staged Spectacle Discloses a Movements Real Values More Than Its Stated Ones]] — same family, and this is the positive-instance version where those are negative. 2. **"A Notice Channel With a 90% Rejection Rate Is Serving a Purpose Other Than Its Stated One"** — atom 2. Genuinely uncaptured and generalizes to any reporting mechanism (DMCA, community flagging, bug bounty intake). Note the tie to the bug-bounty scope change in this same batch: a channel overwhelmed by low-value reports, handled by narrowing rather than by penalizing. 3. Atom 1 is strong but likely better as Evidence in a governance note than as its own Permanent. **Recommendation:** candidate 1 on the next synthesis pass; candidate 2 held pending a second domain instance, which the bug-bounty material may already supply. ### Fleeting Queue - Reconcile the EU figures against the DSA transparency database — a rare externally-verifiable slice of a corporate transparency report. - No compliance rate for ~24,000 government-targeted sites. Volume without outcome is the report's defining omission. - "Provided the page isn't geoblocked for you" — a censorship record that can itself be censored. - 7%/3% trademark rejection: the most citable number in the batch for takedown-abuse arguments. - Platform choice is jurisdiction choice. Strongest concrete argument for owned infrastructure I have read. - Title says Jan–June, URL says Jan–July. Check before citing. <!-- deep-read:sections end --> 📄 **Full text:** [[Full Document Contents/Articles/Transparency Report Update January – June 2026.md|Transparency Report Update: January – June 2026 by Automattic]] ## Highlights - Some reporters carry greater power than others, and government agencies hold the most. Many governments can cut off a platform’s access to their citizens. So, when the government is the reporter, we have to balance user expression with citizen access to the platform. Government power affects both. In the past few years, more and more governments have been reporting content. Between July 1, 2023 and June 30, 2026, [government agencies targeted 17,398 Tumblr sites](https://transparency.automattic.com/transparency-report/tumblr-transparency-report-jan-jun-2026/#government-takedown-demands). On WordPress.com a total of [7,025 sites have been targeted](https://transparency.automattic.com/transparency-report/wordpress-com-transparency-report-jan-jun-2026/#government-takedown-demands) by government reports. ([View Highlight](https://read.readwise.io/read/01m1sgg8ndj5yshsc3bgtjd0hp)) - Tags: 👻 ai highlighted - The transparency of the #MeToo movement unmasked a deeply broken system. Unfortunately, this is the type of content where reporters could cite defamation laws to try and have claims taken down. Imagine that during the #MeToo era, a post alleged a German resident committed sexual assault, and the resident hired a lawyer to write a strongly worded letter in legalese claiming it the content is defamation. ([View Highlight](https://read.readwise.io/read/01m1sgg8nxhyx360wxfbjp3tbf)) - Tags: 👻 ai highlighted