A Right to Copy.  ## Metadata - Author: [[ethanmarcotte.com]] - Full Title: A Right to Copy. - Category: #articles - Document Note: --- tags: [AI, copyright, lawsuits] author: Ethan Marcotte source: ethanmarcotte.com reference: [https://ethanmarcotte.com/wrote/a-right-to-copy/] date created: 2025-10-07 00:00 --- # 🏷️ Title: A right to copy. ## 🧭 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) ### 🔗 References - Include links cited in or relevant to the article. ## 📝 Summary - What happened. - The author attempted to file a claim related to Anthropic’s copyright lawsuit and couldn’t complete it. - Why it matters. - Highlights how hard it is for creators to seek redress or clarity about AI training on their work. - Key stakeholders. - Independent creators, Anthropic, courts, policymakers, AI vendors. - Risks and tradeoffs. - Burdensome processes deter claims; overbroad rights could chill innovation; weak rights erode creator trust. - Bottom line. - The current pathways for asserting rights against AI training are confusing and impractical for individuals. ## 📖 Main Content - Core arguments - Individuals lack clear, usable mechanisms to contest AI training use of their works. - Even when a process exists, it’s opaque, legally fraught, and easy to fail. - Context and terms - Copyright claims: assertions that protected works were used without authorization. - Training data: material ingested to train AI models. - Uncertainties - Scope of fair use in AI training remains unsettled in U.S. courts. - What evidence creators must provide is evolving. ### 🗒️ Notes - “This is one author’s short, failed journey to file a claim in Anthropic’s copyright lawsuit. (Oops. Spoilers, sorry.)” - Describes step-by-step attempts, unclear eligibility, and confusing instructions. - Emphasizes the mismatch between platform assurances and practical redress. - Reflects creator anxiety over unconsented use and the cost of pursuing remedies. ## 📚 References - U.S. Copyright Office: Artificial Intelligence Initiative (https://www.copyright.gov/ai/) - Spawning’s “Have I Been Trained?” (artist data presence) (https://haveibeentrained.com/) - EFF on fair use and AI training overview (https://www.eff.org/deeplinks/2023/12/ai-training-and-fair-use-qa) ## 🗣️ Paraphrase - Paragraph A paraphrase. - I tried to submit a claim tied to Anthropic’s case and hit walls. - Paragraph B paraphrase. - The system’s friction suggests individual creators aren’t really meant to succeed here. ## 📰 Newsletter - AI meets copyright: a creator tried to lodge a claim in Anthropic’s lawsuit and couldn’t. If you publish on the open web, this piece is a reality check on how murky and user-unfriendly redress is today. For WordPress folks: audit your site’s licensing signals, add machine-readable permissions, and review your robots and AI opt-out headers. Helpful reads: the U.S. Copyright Office’s AI initiative and Spawning’s “Have I Been Trained?” to see if your work appears in datasets. ## ✅ Action Items #### DevRel - Publish a plain-language guide on AI training notices and opt-outs; sample headers and robots rules. - Provide code samples for provenance/attribution metadata in SDK docs. - Run office hours on IP questions with legal-approved guardrails. #### OSPO - Define policy on model training with third-party code/content; document acceptable licenses. - Establish scanning/review for dataset provenance; cadence quarterly. - Create an embargo and disclosure flow for IP complaints. #### Community - Host a “Know Your Rights” clinic for creators; share template notices. - Maintain a moderated FAQ on dataset removals and evidence collection. - Sponsor tools that help creators detect model use of their works. #### WordPress - Core: explore canonical support for AI opt-out meta. - Documentation: add examples-first patterns for licensing banners. - Plugins: encourage provenance/proof-of-origin tooling and audits. ## 🧪 Hypotheses - Streamlined, well-documented claim flows increase completion rates by ≥50%. - Machine-readable licensing signals reduce unauthorized dataset inclusion by ≥20%. - Transparency reports from AI vendors improve creator trust scores by ≥30%. - Providing evidence templates raises successful claim submissions by ≥25%. ### 📊 Methodology - A/B test claim portal UX with creators; measure completion and satisfaction. - Crawl sites with/without licensing metadata; compare presence in known datasets. - Survey creators pre/post vendor transparency reports; analyze trust deltas. - Field experiment: distribute evidence checklists; track submission outcomes. ### ❓ Questions - What minimum evidence should creators collect to substantiate a claim? - How should fair use be bounded for large-scale training? - What due process should vendors provide for contested removals? ### 🔮 Future Research - Standardized schema for AI-usage permissions and provenance. - Low-cost verification services for creator identity and ownership. - Auditable logs of dataset composition and updates. ### 💡 Implications - Practice: add opt-out headers and clear licenses. - Policy: mandate transparent claim processes and audit trails. - Security: protect provenance metadata from stripping. - Education: creator toolkits on rights and remedies. ### 🧠 Learning Styles Assessment - Reading level estimate: Flesch-Kincaid Grade ~9–10. #### 👀 Visual learners - Present: none explicit. - Improve: flowchart of the claims process. #### 🔊 Aural learners - Present: none. - Improve: 3-minute audio summary. #### 🗣️ Verbal learners - Present: narrative examples and reflections. - Improve: glossary for legal terms. #### 🧩 Kinesthetic learners - Present: none. - Improve: step-by-step claim checklist. #### 🧠 Logical learners - Present: causal chain from policy to friction. - Improve: decision tree for claim eligibility. #### 🧑‍🤝‍🧑 Social learners - Present: discussion prompts for creator forums. - Improve: peer review of evidence packets. #### 🧘‍♀️ Solitary learners - Present: reflective takeaways. - Improve: self-check quiz on options. #### 🌿 Naturalistic learners - Present: none. - Improve: ecosystem view of creators–vendors–courts. ### 📊 Accessibility Assessment #### Screen reader semantics - Improve: ensure headings and landmarks for any claim portal UI examples. #### Keyboard navigation - Improve: demonstrate tab order and visible focus in sample flows. #### Contrast and readability - Improve: maintain ≥4.5:1 in example screenshots. #### Text resizing and reflow - Improve: responsive examples; respect reduced motion. #### Media and captions - Improve: add captions/transcripts to any demos. #### Forms and interactions - Improve: inline validation and clear error recovery. #### Standards reference - Map gaps to WCAG 2.2 AA: 3.3.1–3.3.3, 2.4.3–2.4.7, 1.4.x. ### 🌍 DEIB Assessment #### Cultural breadth - Improve: non-U.S. legal contexts and translations. #### Inclusive language - Present: plain tone; minimal jargon. - Improve: add definitions for legal terms. #### Neurodiversity support - Improve: chunk steps; consistent patterns. #### Representation - Improve: examples from varied creator communities. #### Psychological safety - Present: respectful framing. - Improve: clear escalation and support resources. ### ✅ Logical Validity #### Strong reasoning - Inductive: firsthand experience illustrates systemic friction; persuasive as a case study. #### Fallacies - Potential hasty generalization: one failed attempt may not represent all systems; mitigate with comparative data. #### Tone of voice - Reflective, dryly humorous, mildly exasperated. ### 🛡️ DARVO Analysis - Deny: Not applicable; author doesn’t deny harm. - Summary: This is one author’s short, failed journey to file a claim in Anthropic’s copyright lawsuit. (Oops. Spoilers, sorry.) - URL: https://ethanmarcotte.com/wrote/a-right-to-copy/ I was keenly watching updates on [the Anthropic copyright lawsuit](https://www.reuters.com/legal/litigation/anthropic-says-chatbot-ai-training-makes-fair-use-books-2025-03-28/), in part because *every single book* I’d ever written appeared in [LibGen, the database of pirated books](https://www.theatlantic.com/technology/archive/2025/03/search-libgen-data-set/682094/) at the heart of the case. When news broke that there was [a frankly astonishing settlement in the case](https://apnews.com/article/anthropic-authors-copyright-judge-artificial-intelligence-9643064e847a5e88ef6ee8b620b3a44c), I figured I might be eligible to file a claim for the affected books. I wasn’t expecting a grand payout, mind; it was really about the principle of the thing. However, when I went to [the new site created for the settlement](https://www.anthropiccopyrightsettlement.com/), I was surprised to see that only *some* of my books appeared when I searched for my name in [the database of affected works](https://secure.anthropiccopyrightsettlement.com/lookup). A number of much, much older books I’d coauthored were listed; none of [the books I wrote myself](https://ethanmarcotte.com/books/) appeared when I looked up my name. I spent some time poking around the claims site’s [labyrinthine FAQs section](https://www.anthropiccopyrightsettlement.com/faq), it seems like a claim can only be filed if a book meets the following criteria: 1. The book was published prior to August 2021; and 2. The book was registered with the [U.S. Copyright Office](https://www.copyright.gov/) within five years of publication, *and* registered before Anthropic downloaded the pirated database in July 2021. [You Deserve a Tech Union](https://ethanmarcotte.com/books/you-deserve-a-tech-union/) was first published in 2023, so it’s immediately ruled out. But my other two books meet the first criterion: [Responsive Design](https://ethanmarcotte.com/books/responsive-web-design/) was published in 2011, with a second edition released in 2014; [Responsive Design: Patterns & Principles](https://ethanmarcotte.com/books/responsive-design-patterns-and-principles/) was published in 2015. So far, so good. But since *none* of my books appear when I search the claims site’s database, I’m guessing the second criterion is where things fall down: namely, I don’t think [my former publisher](https://ethanmarcotte.com/wrote/reacquired/) formally registered my books with the Copyright Office. It’s quite possible they never registered *any* of the books they published: when I search for my former publisher’s name on the claims site, I get zero results. This is where I started to get frustrated. I mean, here’s the thing: my books are copyrighted. They became copyrighted the minute they were published, and that applies to everyone who wrote a book for my former publisher. Heck, the Copyright Office itself says that [registration is voluntary](https://www.copyright.gov/help/faq/faq-general.html). The fact that copyrighted works aren’t eligible *unless* they’ve formally registered feels like a lawyerly dodge: a legal maneuver to help Anthropic reduce the size of its payout pool, and thereby keeping the company from getting rendered down into its component parts during bankruptcy proceedings. The other thing that’s frustrating about this is that I have zero problem with my books getting pirated in the first place. I have *every* problem with my books being used to line the pocket of a few billionaires who don’t care about creators, copyright, or even basic consent. I have *every* problem with my books getting used to prop up a half-baked technology that is destroying the planet, ruining the internet, and supercharging wealth inequality, all while remaining — and I want to be exceedingly clear about this — [a complete and utter failure](https://ethanmarcotte.com/wrote/against-stocking-frames/)., A Right to Copy., ethanmarcotte.com, 01k708xd04cx5ksr0cz50ejsms.