AI vs. The Law - Ryan Roemer, Nearform - CourtneyR.dev's 2nd Brain

AI vs. The Law - Ryan Roemer, Nearform 🔗 **Source:** [AI vs. The Law - Ryan Roemer, Nearform](https://www.youtube.com/watch?v=G7vQSEJwgek) > [!note]+ Document Notes > tags: [AI Law, Copyright, Open Source, jsconf25] > author: [MISSING] > source: YouTube > reference: [https://www.youtube.com/watch?v=G7vQSEJwgek] > date created: > > # 🏷️ Title: AI vs. The Law - Ryan Roemer, Nearform >  ## 🧰 Action Inbox - [H2] Draft a fair-use explainer for devs — https://www.youtube.com/watch?v=G7vQSEJwgek&t=240 #articleseed - [H3] Add “data ingestion method” checklist to OSPO docs — https://www.youtube.com/watch?v=G7vQSEJwgek&t=372 #xref - [H7] Enable code provenance scanning in CI — https://www.youtube.com/watch?v=G7vQSEJwgek&t=929 #share - [H8] Update contributor guide on human authorship — https://www.youtube.com/watch?v=G7vQSEJwgek&t=1126 #articleseed - [H9] Add prompt-only copyright warning banner — https://www.youtube.com/watch?v=G7vQSEJwgek&t=1148 #addendum - [H10] Set “human-in-the-loop” policy for AI code — https://www.youtube.com/watch?v=G7vQSEJwgek&t=1554 #xref - [H11] Track key AI cases in a living note — https://www.youtube.com/watch?v=G7vQSEJwgek&t=806 #share ## 🧭 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 - US Copyright Office AI guidance and policy page: https://www.copyright.gov/ai/ - Thaler v. Perlmutter opinion (AI authorship not protected): https://www.courtlistener.com/docket/63352999/thaler-v-perlmutter/ *(free access; document availability can vary)* - Overview of Thomson Reuters v. Ross Intelligence docket: https://www.courtlistener.com/docket/17082981/thomson-reuters-enterprises-llc-v-ross-intelligence-inc/ *(access may require navigation; some filings may be paywalled)* - Background on GitHub Copilot litigation: https://www.eff.org/deeplinks/2023/11/appeals-court-lets-github-copilot-lawsuit-proceed *(advocacy summary; monitor for updates)* - Video of the talk with transcript: https://www.youtube.com/watch?v=G7vQSEJwgek ## 📝 Summary - The speaker surveys US copyright law and AI. The talk targets code. See [H1], [H2]. - Fair use remains contested. Courts split on training data legality. See [H11]. - Data ingestion methods shape risk. Contract and scraping issues matter. See [H3]. - AI tools can emit copyrighted code. Provenance controls reduce risk. See [H7]. - US law requires human authorship. Prompts alone do not suffice. See [H8], [H9]. - Bottom line: favor human control, clear policies, and auditing. Cite primary guidance. ## 📖 Main Content - The talk limits scope to US law and code. This narrows variables. See [H1]. - Fair use uses a four-factor test. Purpose, nature, amount, and market harm guide analysis. See [H2]. - Courts disagree on training data. Outcomes vary by source and use. See [H11]. - Data collection methods matter. Scraping, contracts, and piracy shape exposure. See [H3]. - AI coding tools can reproduce copyrighted snippets. Use scanners and guardrails. See [H7]. - Human authorship controls ownership. US guidance rejects purely AI authorship. See [H8]. - Prompts alone lack copyright protection. Selection and arrangement by humans can qualify. See [H9]. - Uncertainty persists. Therefore, implement measured, auditable workflows now. ### 🗒️ Notes #### ⏱️ [00:02:51](https://www.youtube.com/watch?v=G7vQSEJwgek&t=171) Scope and limits ### [H1] Talk focuses on US law and code > [!quote] [AI vs. The Law](https://www.youtube.com/watch?v=G7vQSEJwgek) by an unnamed author via [[YouTube]]. Published [MISSING]. Accessed . > “I'm just going to be talking about US law and pretty much specifically code.” Scope clarity reduces misapplication across jurisdictions. [view highlight](https://www.youtube.com/watch?v=G7vQSEJwgek&t=171) #xref ID:rw_[MISSING] #### ⏱️ [00:04:00](https://www.youtube.com/watch?v=G7vQSEJwgek&t=240) Fair use frame ### [H2] Fair use relies on four factors > [!quote] [AI vs. The Law](https://www.youtube.com/watch?v=G7vQSEJwgek) > “It is a four factor test…” Use this lens when evaluating training and outputs. [view highlight](https://www.youtube.com/watch?v=G7vQSEJwgek&t=240) #articleseed ID:rw_[MISSING] #### ⏱️ [00:06:12](https://www.youtube.com/watch?v=G7vQSEJwgek&t=372) Ingestion risk ### [H3] Data ingestion method affects legality > [!quote] [AI vs. The Law](https://www.youtube.com/watch?v=G7vQSEJwgek) > “does the manner in which you collect and ingest this data actually matter.” Document collection methods and permissions. [view highlight](https://www.youtube.com/watch?v=G7vQSEJwgek&t=372) #xref ID:rw_[MISSING] #### ⏱️ [00:08:02](https://www.youtube.com/watch?v=G7vQSEJwgek&t=482) Early holding ### [H4] Thomson Reuters v. Ross finds infringement > [!quote] [AI vs. The Law](https://www.youtube.com/watch?v=G7vQSEJwgek) > “there was a holding of infringement.” Use competitor materials cautiously; market harm weighs heavily. [view highlight](https://www.youtube.com/watch?v=G7vQSEJwgek&t=482) #xref ID:rw_[MISSING] #### ⏱️ [00:11:45](https://www.youtube.com/watch?v=G7vQSEJwgek&t=705) Split outcomes ### [H5] Cadri v. Meta deems use transformative > [!quote] [AI vs. The Law](https://www.youtube.com/watch?v=G7vQSEJwgek) > “the Cadri court came down and said that use was transformative.” Track case differences before setting policy. [view highlight](https://www.youtube.com/watch?v=G7vQSEJwgek&t=705) #xref ID:rw_[MISSING] #### ⏱️ [00:12:39](https://www.youtube.com/watch?v=G7vQSEJwgek&t=759) Scraping disputes ### [H6] Reddit sues Anthropic over scraping > [!quote] [AI vs. The Law](https://www.youtube.com/watch?v=G7vQSEJwgek) > “Reddit is suing Anthropic for scraping and training on their data.” Expect contract and tort claims beyond copyright. [view highlight](https://www.youtube.com/watch?v=G7vQSEJwgek&t=759) #xref ID:rw_[MISSING] #### ⏱️ [00:15:29](https://www.youtube.com/watch?v=G7vQSEJwgek&t=929) Output risk ### [H7] AI tools can emit copyrighted code > [!quote] [AI vs. The Law](https://www.youtube.com/watch?v=G7vQSEJwgek) > “this is actually producing like copyrighted code” Adopt provenance scanning and snippet filters. [view highlight](https://www.youtube.com/watch?v=G7vQSEJwgek&t=929) #share ID:rw_[MISSING] #### ⏱️ [00:18:43](https://www.youtube.com/watch?v=G7vQSEJwgek&t=1123) Authorship rule ### [H8] US law requires human authorship > [!quote] [AI vs. The Law](https://www.youtube.com/watch?v=G7vQSEJwgek) > “the US copyright act requires human authorship.” Plan rights around human control and edits. [view highlight](https://www.youtube.com/watch?v=G7vQSEJwgek&t=1123) #articleseed ID:rw_[MISSING] #### ⏱️ [00:19:11](https://www.youtube.com/watch?v=G7vQSEJwgek&t=1149) Prompt limits ### [H9] Prompts alone do not secure copyright > [!quote] [AI vs. The Law](https://www.youtube.com/watch?v=G7vQSEJwgek) > “prompts and instructions alone are not sufficient.” Require human selection and arrangement. [view highlight](https://www.youtube.com/watch?v=G7vQSEJwgek&t=1149) #addendum ID:rw_[MISSING] #### ⏱️ [00:25:54](https://www.youtube.com/watch?v=G7vQSEJwgek&t=1554) Caution path ### [H10] Favor human involvement in AI workflows > [!quote] [AI vs. The Law](https://www.youtube.com/watch?v=G7vQSEJwgek) > “favor as much human involvement as possible.” Adopt review gates and edit thresholds. [view highlight](https://www.youtube.com/watch?v=G7vQSEJwgek&t=1554) #xref ID:rw_[MISSING] #### ⏱️ [00:13:26](https://www.youtube.com/watch?v=G7vQSEJwgek&t=806) Legal uncertainty ### [H11] Courts split on AI fair use > [!quote] [AI vs. The Law](https://www.youtube.com/watch?v=G7vQSEJwgek) > “the courts are now split on fair use” Build policies that tolerate variance. [view highlight](https://www.youtube.com/watch?v=G7vQSEJwgek&t=806) #articleseed ID:rw_[MISSING] ## 📚 References - US Copyright Office — AI policy hub: https://www.copyright.gov/ai/ - Thaler v. Perlmutter decision summary: https://www.courtlistener.com/docket/63352999/thaler-v-perlmutter/ - Thomson Reuters v. Ross docket overview: https://www.courtlistener.com/docket/17082981/thomson-reuters-enterprises-llc-v-ross-intelligence-inc/ - EFF on Copilot litigation status: https://www.eff.org/deeplinks/2023/11/appeals-court-lets-github-copilot-lawsuit-proceed - Talk video: https://www.youtube.com/watch?v=G7vQSEJwgek ## 🔖 Zettel Candidates - **Fair use relies on four factors** — Anchor training and output reviews to the test. https://www.youtube.com/watch?v=G7vQSEJwgek&t=240 - **Data ingestion method affects legality** — Track sources, agreements, and robots rules. https://www.youtube.com/watch?v=G7vQSEJwgek&t=372 - **Thomson Reuters v. Ross finds infringement** — Competitor inputs raise market-harm risk. https://www.youtube.com/watch?v=G7vQSEJwgek&t=482 - **Cadri v. Meta deems use transformative** — Expect divergent holdings across courts. https://www.youtube.com/watch?v=G7vQSEJwgek&t=705 - **Reddit sues Anthropic over scraping** — Contract and tort expand liability. https://www.youtube.com/watch?v=G7vQSEJwgek&t=759 - **AI tools can emit copyrighted code** — Add provenance scans and snippet filters. https://www.youtube.com/watch?v=G7vQSEJwgek&t=929 - **US law requires human authorship** — Assign rights through human edits and control. https://www.youtube.com/watch?v=G7vQSEJwgek&t=1123 - **Prompts alone do not secure copyright** — Require selection and arrangement by humans. https://www.youtube.com/watch?v=G7vQSEJwgek&t=1149 - **Favor human involvement in AI workflows** — Define edit thresholds and review steps. https://www.youtube.com/watch?v=G7vQSEJwgek&t=1554 - **Courts split on AI fair use** — Maintain adaptable policy stances. https://www.youtube.com/watch?v=G7vQSEJwgek&t=806 ## 🗣️ Paraphrase - The talk outlines how US copyright intersects with AI training and code. - It stresses human control, careful data sourcing, and pragmatic guardrails. ## 📰 Newsletter - AI law moves fast; your repos should move carefully. - Add provenance scanning and human review to AI-assisted code paths. - Refresh contributor docs with fair-use basics and authorship rules. - Read the US Copyright Office guidance for current positions: https://www.copyright.gov/ai/ - Watch the session: https://www.youtube.com/watch?v=G7vQSEJwgek ## ✅ Action Items #### DevRel - Publish an “AI and Code Provenance” guide with examples. - Run office hours on fair use, authorship, and repo hygiene. - Demo CI checks that flag non-attributable snippets. - Track incident metrics from provenance violations. - Share case updates with engineering channels monthly. #### OSPO - Add a data ingestion register with source and license fields. - Require human-in-the-loop edits for AI outputs above X lines. - Define AI code attribution and third-party notice procedures. - Establish an escalation path to counsel for flagged code. - Maintain a watchlist of active AI cases with short briefs. #### Community - Offer maintainers a provenance scanner starter kit. - Host a “Know your license” clinic with Q&A. - Publish an FAQ on prompts, authorship, and contributor rights. - Encourage sharing vetted model settings and filters. #### WordPress - Core: document AI usage rules in contributing.md. - Core Performance: log provenance in profiling artifacts. - Core AI: publish evaluation sets and redaction rules. - **Core Program Team:** record decisions on AI guardrails. - Playground: ship a provenance demo lab for learners. - Documentation: add examples-first licensing guidance. #### The WPOCC - Fund microgrants for provenance tooling and training. - Sponsor legal clinics for maintainers on AI authorship. - Pilot matched grants for compliance improvements. ### FAIR - Ship operator agreements covering AI use, edits, and signatures. - Provide scorecards for provenance SLOs and release attestations. - Publish anti-capture covenants for data licensing. ## 🧪 Hypotheses - Human edit thresholds reduce copyright risk incidents by 30%. - Provenance scanners lower license-mismatch PRs by 25%. - Ingestion registers cut unknown-source usage by 40%. - Contributor training increases correct attribution by 20%. ### 📊 Methodology - A/B test CI pipelines with and without provenance checks. - Audit PRs for origin, license, and edit depth. - Track violations per KLoC over releases. - Survey contributors on policy clarity quarterly. ### ❓ Questions - What human-edit threshold best balances speed and safety? - Which scanner signals correlate with real legal risk? - How should WordPress teams record AI involvement in code? ### 🔮 Future Research - Compare outcomes across different AI code filters. - Study contract-based scraping cases versus copyright claims. - Explore licensing patterns for model weights and datasets. ### 💡 Implications - Teams should record data sources and model settings. - Projects should require human edits for AI code. - Communities should fund provenance tooling and education. ### 🧠 Learning Styles Assessment - Reading level estimate: Flesch–Kincaid Grade ≈ 10.5. #### 👀 Visual learners - Present: slides and code snippets. - Improve: flowcharts of fair-use decisions and workflows. #### 🔊 Aural learners - Present: recorded talk with captions. - Improve: 3-minute audio summary with key takeaways. #### 🗣️ Verbal learners - Present: definitions and examples. - Improve: glossary for legal and licensing terms. #### 🧩 Kinesthetic learners - Present: CLI demo concept. - Improve: hands-on provenance labs and checklists. #### 🧠 Logical learners - Present: factor tests and comparisons. - Improve: decision trees and risk matrices. #### 🧑‍🤝‍🧑 Social learners - Present: Q&A prompts. - Improve: peer review rubrics for AI PRs. #### 🧘‍♀️ Solitary learners - Present: transcript and slides. - Improve: self-check quizzes on scenarios. #### 🌿 Naturalistic learners - Present: none. - Improve: discuss compute impacts and green policies. ### 📊 Accessibility Assessment #### Screen reader semantics - Present: clear headings. - Improve: alt text and transcript structuring in shared slides. #### Keyboard navigation - Present: standard player controls. - Improve: documented shortcuts and skip links. #### Contrast and readability - Present: variable slide contrast. - Improve: enforce 4.5:1 and adequate line height. #### Text resizing and reflow - Present: responsive video player. - Improve: verify 200% zoom on shared docs. #### Media and captions - Present: YouTube captions available. - Improve: downloadable transcript and audio description. #### Forms and interactions - Present: N/A. - Improve: accessible feedback forms and toasts. #### Standards reference - Map improvements to WCAG 2.2 AA. ### 🌍 DEIB Assessment #### Cultural breadth - Present: US-centric legal framing. - Improve: add non-US perspectives and links. #### Inclusive language - Present: neutral legal terms. - Improve: plain-language summaries for newcomers. #### Neurodiversity support - Present: structured segments. - Improve: timers and visual anchors for concepts. #### Representation - Present: single speaker format. - Improve: link diverse legal experts and maintainers. #### Psychological safety - Present: cautionary framing without blame. - Improve: clear escalation paths for concerns. ### ✅ Logical Validity #### Strong reasoning - The four-factor test summary matches statute ([H2]). It clarifies analysis. - Human authorship rule aligns with Thaler v. Perlmutter ([H8]). It grounds policy. - Data-ingestion emphasis recognizes contract and tort exposure ([H3]). It reduces surprises. #### Fallacies - Overgeneralizing split rulings to all cases risks hasty generalization. Note bounds ([H11]). - Assuming scanners catch all risk invites false security. Add human review ([H7]). #### Tone of voice - The talk uses pragmatic, cautionary, engineering-friendly language. It avoids legal absolutism. ### 🛡️ DARVO Analysis - Deny: none observed. - Attack: none observed. - Reverse victim/offender: none observed. - The rhetoric favors shared responsibility and risk reduction. ## 🏷️ Tag Coverage #articleseed: 3 • #xref: 4 • #share: 2 • #bmf: 0 • #addendum: 1 ## 🎯 Conclusion - Treat AI legal alignment as operational work, not vibes. - Record sources, require human edits, and scan outputs. - Track cases and update policies with counsel and community. ## 🧱 Refactor Appendix (append after 🎯 Conclusion) ### 📚 Literature Split Pad - Purpose: Stage Konik CEE units that become per-source Literature notes. - Split level: **H3**. - Include verbatim quotes under 25 words and a [view highlight] link. ### [H1] Talk focuses on US law and code > [!quote] [AI vs. The Law - Ryan Roemer, Nearform](https://www.youtube.com/watch?v=G7vQSEJwgek) by an unnamed author via [[YouTube]]. Published [MISSING]. Accessed . > “I'm just going to be talking about US law and pretty much specifically code.” Two lines: scope control avoids jurisdiction creep. [view highlight](https://www.youtube.com/watch?v=G7vQSEJwgek#H1) #xref #articleseed ID:rw_[MISSING] ### [H2] Fair use relies on four factors > [!quote] [AI vs. The Law](https://www.youtube.com/watch?v=G7vQSEJwgek) > “It is a four factor test…” Two lines: use as review checklist for training and outputs. [view highlight](https://www.youtube.com/watch?v=G7vQSEJwgek#H2) #xref ID:rw_[MISSING] --- ### 🔖 Atomic Split Pad (Zettel Candidates) - Purpose: Promote reusable claims into Atomic notes. - Split level: **H4**. - Each item: one idea, present tense, specific, testable. #### [Z1] Data ingestion method affects legality - Applies when scraping or licensing third-party sources. - Pointer: talk at t=372. - Links: [[AI-Compliance]] [[Data-Governance]] #### [Z2] Prompts alone do not secure copyright - Applies to code and media outputs. - Pointer: talk at t=1149; USCO AI guidance. - Links: [[Authorship]] [[Licensing]] --- ### ✏️ Fleeting Queue (Extract by Selection) - Purpose: Capture quick actions or sparks for Fleeting notes. - Extraction: **Select one line** → “Refactor: Extract selection to new note (content only).” - Keep each line short and self-contained. - [F1] Write a page on “human edit thresholds” — context/https://www.youtube.com/watch?v=G7vQSEJwgek #fleeting #inbox - [F2] Map provenance scanner options for WordPress teams — context/https://www.youtube.com/watch?v=G7vQSEJwgek #fleeting #inbox - [F3] Build a living case tracker template — context/https://www.youtube.com/watch?v=G7vQSEJwgek #fleeting #inbox ---