The pattern in the public record.

On January 3, 2023, the New York City Department of Education blocked access to ChatGPT on all DOE devices and networks. The official rationale, from spokesperson Jenna Lyle:

“Due to concerns about negative impacts on student learning, and concerns regarding the safety and accuracy of content, access to ChatGPT is restricted on New York City Public Schools' networks and devices.”

— Jenna Lyle, NYC DOE spokesperson, January 2023

“While the tool may be able to provide quick and easy answers to questions, it does not build critical-thinking and problem-solving skills, which are essential for academic and lifelong success.”

— Jenna Lyle, NYC DOE spokesperson, January 2023

Four-and-a-half months later, on May 18, 2023, Chancellor David C. Banks published an op-ed in Chalkbeat reversing the ban. The piece was titled ChatGPT caught NYC schools off guard. Now, we're determined to embrace its potential. The key admission:

“The knee-jerk fear and risk overlooked the potential of generative AI to support students and teachers, as well as the reality that our students are participating in and will work in a world where understanding generative AI is crucial.”

— Chancellor David C. Banks, May 18, 2023

Why this case is the gold standard: §7's pathology is hard to demonstrate when institutions never publicly walk back the decline. Most refusals-to-decompose stay in place. NYC DOE's reversal — and the explicit naming of theknee-jerk reaction by the Chancellor in his own voice — is unusually direct evidence that the pattern is recognizable enough to be admitted. The original announcement came from a spokesperson; the reversal came from the Chancellor. The asymmetry itself is data.

Sciences Po in Paris announced in late January 2023 that students were forbidden from using ChatGPT or other AI-based tools for any written work or presentations without transparent referencing, with non-citation penalized as academic fraud. The framing was less of a blanket ban than NYC DOE's — supervisor-approved use was allowed for specific course purposes — but the institutional reflex was the same: react first, decompose later. Sciences Po was one of the first major European universities to issue a formal policy, and the policy has since been refined as the discipline-specific implications became clearer.

The case illustrates a softer variant of the §7 pattern: the institution does not invoke the public in the same way NYC DOE did, but it still substitutes blanket prohibition for the harder work of decomposition. Notably, Sciences Po did not issue a similar public reversal — the policy continues to evolve through internal academic governance rather than through op-ed admission.

xAI's Colossus 1 data center in South Memphis became the subject of sustained documented opposition through 2024–2026. Key facts in the public record:

• ·Originally promised an on-site wastewater recycling facility to avoid drawing from the Memphis Sand Aquifer. That commitment was paused indefinitely. Expected water demand: approximately 5+ million gallons per day, in a region where arsenic contamination already threatens drinking water supply.
• ·Colossus 1 sits in South Memphis, a predominantly Black neighborhood where residents face cancer risk approximately four times the national average. Thermal imagery has documented over 30 unpermitted natural gas turbines operating at the site.
• ·April 2026: the NAACP filed a federal lawsuit alleging that the unpermitted turbines release harmful pollutants and disproportionately impact predominantly Black neighborhoods.
• ·April 2026: U.S. Senator Sheldon Whitehouse (Senate Environment & Public Works Committee, minority) opened a formal inquiry into xAI's pattern of operating illegal data center gas plants.

This is not the §7 mirror failure. Local water-protection groups, environmental-justice coalitions, and federal officials are doing exactly the analytic work the framework asks for — identifying specific actors, specific permits bypassed, specific air-quality impacts, specific health outcomes in specific communities. The decomposition has been done. The opposition stands.

Why this matters for the framework: a constituency that has read this record correctly may be reluctant to embrace anything labeled AI in this market — even tools that are clearly distinguishable from frontier-LLM training infrastructure (locally-served, no hyperscaler in the loop, commodity hardware). Recognizing this as constituency-management in a hot political moment, not as refusal-to-decompose, matters for §7's honest reading and for the framework's broader relational strategy. The right response to a substantive-opposition context is not to push the decomposition harder; it is to understand why association costs are high and to wait for the local context to shift.

The 2023 Hollywood strikes produced contracts that decompose AI into operationally distinct sub-cases. Each agreement addresses different categories with different protections.

WGA (Writers Guild of America)

• ·AI is not a writer. No form of AI (generative or otherwise) may be considered a writer for contractual purposes.
• ·Material produced by AI cannot be considered literary material.
• ·Writers cannot be cut from the creative process or paid less because of AI involvement.
• ·The agreement reserves WGA's right to assert that use of literary materials for training generative AI violates the agreement or applicable law.

SAG-AFTRA (Screen Actors Guild · AFTRA)

• ·Digital Replicas — creating an Employment-Based Digital Replica requires producers to give actors 48 hours notice, obtain clear and conspicuous consent, and pay for each specific use.
• ·Digital Alterations — substantive changes to recorded performance require explicit consent.
• ·Synthetic Performers — completely artificial human-appearing performers require producers to notify SAG-AFTRA and bargain in good faith over consideration.

Why this matters for the framework: both contracts decompose the AI category into specific operational sub-cases — exactly what §4 argues for. The decomposition was achieved through two separate ~148-day strikes, with substantial union legal and research staff producing the contractual language. Institutional capacity for §4-level decomposition is real but expensive. Smaller institutions — most mission-aligned nonprofits, most public-school districts, most municipal governments — do not have this capacity. The §7 mirror failure is so common partly because institutions without SAG-AFTRA-scale resources cannot afford to do the decomposition themselves and reach for the easier blanket refusal instead.

The American Library Association did not ban ChatGPT or any other generative AI tool. Instead it formed working groups (including Core's Artificial Intelligence and Machine Learning in Libraries Interest Group) and produced practical guidance materials. The Association of College and Research Libraries (ACRL) approved formal AI Competencies for Academic Library Workers in October 2025. The ALA's posture emphasizes domain-specific decomposition: information integrity, copyright, privacy, misinformation, and equitable access as distinct issues to be analyzed separately rather than collapsed into a single AI category.

Why this is a useful counter-example: not every institution refuses to decompose. The structural conditions under which ALA succeeded — membership-driven mandate, professional-society infrastructure for slow deliberation, low political cost of being seen to engage technically with new tools — are themselves worth analyzing. They suggest where the framework's engagement strategy can productively focus, and where it is structurally unlikely to land.