by RALPH, Research Fellow, Recursive Institute Adversarial multi-agent pipeline · Institute-reviewed. Original research and framework by Tyler Maddox, Principal Investigator.
The Founding Pattern: Edison’s Trust and the Hollywood Bypass
In December 1908, Thomas Edison’s Motion Picture Patents Company consolidated monopoly control over American filmmaking through a simple, elegant mechanism: patent licensing over cameras, projectors, and an exclusive supply agreement with Eastman Kodak for raw film stock. If you wanted to make a movie, you needed Edison’s permission. If you wanted to show one, you paid Edison’s royalties. The Trust filed over 289 patent lawsuits against independents and employed detectives to physically confiscate unlicensed equipment.
The independents did not defeat it. They left.
Carl Laemmle, William Fox, and Adolph Zukor moved production three thousand miles west to Southern California. The geographic distance created enforcement lag. Local courts proved less sympathetic to East Coast patent monopolies. By the time the federal courts ruled the Trust illegal in 1915, the independents had already built the studio system. The Trust member companies — Essanay, Kalem, Lubin, Selig, even Edison’s own studio — collapsed between 1915 and 1918. Vitagraph, the last survivor, was absorbed by Warner Brothers in 1925.
This is the founding pattern for what this framework calls entity substitution: institutional protections die not through direct legal defeat, but through the economic death or bypass of the entities that carry them. The protection survives on paper. The entity carrying it does not survive in the market. The result is the same as repeal, but without the political fight. [Framework — Original]
The Edison case matters because it reveals the mechanism in its purest form. The Trust’s patents were valid. Its enforcement was real. The independents’ response was not to challenge the patents but to make them irrelevant by operating where they could not be enforced.
What makes the current moment different is that AI does not require geographic relocation to achieve the bypass. The “Hollywood” that new entrants are moving to is not a place. It is a cost structure.
The academic framework is well-established. A 2023 paper in the Journal of Macromarketing proposed a general theory of regulatory arbitrage applicable across film, finance, drugs, and labor markets. Kathleen Thelen and Paul Pierson’s work on institutional change identifies the specific vulnerability: institutions that depend on ongoing participation by specific economic actors are subject to “drift” — the rules remain formally unchanged but lose effectiveness as the context shifts around them. The institutional form survives. The institutional function does not. [Measured]
The WGA/SAG Case: Winning Protections That Attach to Dying Entities
The 2023 WGA and SAG-AFTRA strikes were, by any reasonable measure, a labor victory. The Writers Guild secured provisions establishing that AI cannot write or rewrite literary material, that AI-generated content cannot be used to deny writer credits, that companies must disclose AI-generated material to writers, and that minimum staffing requirements prevent replacement of writers’ rooms by AI systems. SAG-AFTRA secured consent requirements for digital replicas [1].
These protections are real. They are binding on the signatories. And therein lies the problem.
The signatories are the members of the Alliance of Motion Picture and Television Producers — over 350 producers, streamers, studios, and networks including Disney, Netflix, Amazon, Apple, Warner Bros. Discovery, Paramount, Sony, and Fox. The protections do not bind entities that never signed the agreements. YouTube-native creators fall outside guild coverage entirely.
The question is not whether the protections are good. The question is whether the entities carrying those protections can survive the cost differential against competitors who never assumed them.
March 2026 update — the 2026 negotiations. The current WGA and SAG-AFTRA contract negotiations, with the WGA contract expiring May 1, 2026, are testing entity substitution under intensified conditions [2]. SAG-AFTRA’s most prominent proposal is a “Tilly tax” — a fee required whenever a studio opts to use an AI performer instead of a human actor [3]. This represents an attempt to convert entity-dependent protections into quasi-activity-dependent ones: rather than simply binding signatories, it would impose costs on the activity of using AI in production, regardless of which entity performs it. Whether the AMPTP accepts such a provision — and whether it survives the cost pressure that entity substitution creates — is the single most important near-term test of the framework.
The financial evidence has deteriorated since the 2023 strikes. Paramount Skydance’s debt has been downgraded to junk status (BB+) by Fitch Ratings following its $110 billion acquisition of Warner Bros. Discovery [4]. S&P Global placed the combined entity on negative credit watch, with adjusted leverage at 4.8x and potential to exceed 7x — far above the 4.5x threshold needed to maintain even the current BB+ rating [5]. The combined entity carries approximately $79 billion in long-term debt [6]. Paramount Skydance will pay WBD’s shareholders a daily ticking fee starting after September 30, 2026, which could add $650 million per quarter in additional costs [4]. This is not a company in a position to absorb substantial new labor obligations.
Meanwhile, a typical scripted television episode costs approximately $9 million under guild agreements. Premium shows run much higher. SAG-AFTRA day rates are $1,246; weekly rates for principals on hour-long shows run to nearly $11,000. Pension and health contributions add 23.5% on top. McKinsey estimates that AI adoption could reduce production costs by 30% overall for content creators, with 70-90% reductions possible for high-volume or routine content. [Estimated] The cost differential between guild-obligated production and AI-native production is not marginal. It is structural.
The enforcement landscape already shows the cracks. In May 2025, SAG-AFTRA filed an unfair labor practice charge against Llama Productions for using an AI-generated voice replica of the late James Earl Jones as Darth Vader in Fortnite — the first major enforcement action against AI voice cloning of a deceased performer. The 2024-2025 SAG-AFTRA video game strike centered on producers’ refusal to extend AI protections to motion capture artists. Formosa Interactive was accused of evading the strike by transferring a title to a shell company and posting “non-union talent only” casting calls. The shell company bypass, the selective refusal to extend protections, the exploitation of coverage gaps — these are not violations of the 2023 contracts. They are demonstrations that the boundaries of those contracts create exploitable space.
Iron Lung: The Proof of Concept
In January 2026, Mark Fischbach — known online as Markiplier — released Iron Lung, a feature film he wrote, directed, starred in, and self-financed for approximately $3 million. Fan demand expanded his initial plan of 50-100 theaters to roughly 3,000, including all three major chains. As of mid-February 2026, the film had grossed in the range of $30-35 million worldwide — a return on investment exceeding ten times its budget. [Measured]
Iron Lung was not AI-generated. Fischbach made a real movie with real actors. The point is not that AI replaced anything in this specific case. The point is that the production and distribution infrastructure that guild protections were designed to regulate — the studio system — is no longer the only viable path to commercial filmmaking. The bypass exists. The question is what happens when AI tools make it available not just to creators with 37 million subscribers but to anyone with a laptop and a subscription.
The trajectory is visible. YouTube CEO Neal Mohan’s 2026 positioning frames creators as “the new stars and studios” — not metaphorically, but structurally. New entities like Further Adventures are explicitly designed to turn YouTube-native storytelling into feature film IP. The creator-as-studio model is being systematized, and these studios operate outside the guild framework.
The tools available to a solo filmmaker in 2026 make the one-person studio a practical reality. Synthesia and DeepBrain AI generate hyper-realistic avatars with facial expressions and lip-sync across 80 languages. ElevenLabs provides voice cloning increasingly adopted across media production. LTX Studio offers a browser-based script-to-screen pipeline. Runway Gen-4.5 handles AI video editing with exports up to 4K resolution. The remaining gap — perfectly consistent characters across long-form narrative — is closing rapidly.
The Dissolution Mechanism: Section 1113 and the Bankruptcy Path
Understanding how guild protections actually dissolve requires understanding bankruptcy law. Section 1113 of the U.S. Bankruptcy Code permits rejection of collective bargaining agreements in Chapter 11 proceedings if the debtor can demonstrate that the union refused without good cause a proposal that satisfies statutory requirements, and that the balance of equities clearly favors rejection.
This is not a theoretical mechanism. It has been exercised repeatedly across American industry.
Continental Airlines filed Chapter 11 in September 1983 — while still solvent, with $58 million in cash reserves — specifically to break its union contracts. The move was so aggressive that Congress added Section 1113 in 1984 to require good-faith negotiations before rejection. But the procedural safeguard did not prevent the pattern from repeating. TWA went through bankruptcy three times before American Airlines acquired it in 2001. Bethlehem Steel dumped its pension liabilities onto the Pension Benefit Guaranty Corporation and cut off retiree health insurance. Delphi’s CEO Steve Miller, who had previously presided over Bethlehem Steel’s restructuring, replicated the same playbook and stated publicly: “If you can’t get an agreement, you go back to the court and ask it to reject the contract. Then it’s a free-for-all.” [Measured]
In entertainment, residual and royalty obligations may face an even weaker position than collective bargaining agreements. The Third Circuit has ruled that certain royalty claims constitute unsecured creditor claims whose future payments can be discharged in bankruptcy — though the specific treatment depends on contract structure, security interests, and the jurisdiction. Unlike CBAs, which receive the procedural protections of Section 1113, residuals are generally treated as unsecured claims, placing them behind secured creditors and priority claimants in the distribution waterfall.
The SAG-AFTRA 2023 contract secured $317.2 million in pension and health contributions and $697.6 million in wages and residuals over its term. These obligations are binding on signatory entities. They are also claims that would be subject to discharge if those entities entered Chapter 11. The protection is only as durable as the entity’s balance sheet.
The Paramount Skydance situation illustrates how merger activity can accelerate the dissolution mechanism. When Paramount Skydance announced its $110 billion acquisition of Warner Bros. Discovery [6], the combined entity’s debt load nearly doubled. The Skydance investor group projects the combined company “to be investment-grade by all rating agencies sometime in 2026,” with the debt-to-operating-income ratio declining from approximately 4.3x to 2.4x by 2027 [4]. But this projection depends on aggressive cost synergies and revenue growth in an environment where linear television revenue is declining 7% annually and streaming economics remain challenging. If the synergy targets are not met — and the base rate for achieving projected merger synergies in media is not encouraging — the combined entity faces a debt maturity wall with reduced cash flow, creating precisely the conditions under which Section 1113 becomes relevant. The ticking fee alone — $650 million per quarter after September 2026 — represents a structural cash drain that has no precedent in entertainment industry mergers [4].
The General Pattern: Where Entity Substitution Has Already Happened
Entertainment is not the first industry to face this dynamic.
Trucking. In the 1970s, approximately 80% of the trucking industry was unionized. The Teamsters represented over two million truck drivers. Then the Motor Carrier Act of 1980 deregulated the freight industry. By 1999, union density had fallen to 20%. Today, fewer than 60,000 Teamsters freight workers remain. Yellow Corporation, once the largest unionized LTL carrier, ceased all operations in July 2023 and filed Chapter 11 in August, eliminating 30,000 jobs — 22,000 of them Teamster-represented. FedEx Ground and Amazon Logistics, the entities that replaced Yellow’s market share, never assumed comparable union obligations. [Measured]
Journalism. Newspaper guild contracts created real protections — staffing minimums, severance requirements, editorial independence provisions. When Alden Global Capital acquires a newspaper, it cuts staff dramatically, sells real estate, outsources production. BuzzFeed News shut down entirely in April 2023. Vice went bankrupt. The entity substitution cycle completed within a decade: legacy entities with protections die, replacements that briefly unionize prove economically unviable, the protections dissolve with both. [Measured]
Manufacturing. The shift from unionized domestic production to non-union overseas supply chains reduced manufacturing unionization from a majority to 5.9% in 2024 — losing 167,000 union members since 2019 alone. More than 4.7 million manufacturing jobs have been lost since January 2000. [Measured]
The pattern is remarkably consistent. First, a regulatory framework creates obligations for incumbent entities. Second, a structural change enables new entrants to perform the same function without those obligations. Third, the cost differential widens until incumbents cannot compete. Fourth, the incumbents restructure, merge, or enter bankruptcy. Fifth, the replacement entities continue operating without the protections.
What varies is the speed. Trucking deregulation took decades to complete the cycle. Journalism’s transition compressed into roughly fifteen years. The question for entertainment is whether AI acceleration compresses the timeline further — whether the cost curve bends fast enough that studios carrying tens of billions in debt alongside substantial guild obligations face existential pressure within a single contract cycle rather than over a generation.
The Aggregate Demand Crisis (MECH-010) connects here through a specific channel. When legacy entities carrying labor obligations dissolve, the wages, benefits, residuals, and pension contributions they funded disappear from the demand circuit. Entity substitution is a specific channel of demand destruction — every dollar of guild wages that goes unpaid when a studio enters Chapter 11 is a dollar removed from the consumer economy. The SAG-AFTRA pension and health fund depends on continued contributions from signatory employers. If those employers shrink or restructure, the fund’s actuarial assumptions collapse — not because the contract was violated, but because the entities funding it can no longer sustain their share. This is the Ratchet-Demand Death Spiral operating through the specific channel of institutional dissolution.
The AI companies that will supply the tools for non-guild production operate with a fundamentally different labor structure. No successful unionization has been reported at OpenAI, Anthropic, or Google DeepMind. Their extended workforce — contractors, gig workers, data annotators — comprises 30-50% of total headcount at many AI firms, structured specifically to avoid the benefits obligations (health insurance, pension contributions, unemployment insurance) that guild contracts mandate. The entities replacing studio production do not just lack guild agreements. They are structurally organized to avoid comparable labor obligations at every level.
The Critical Distinction: Entity-Dependent vs. Activity-Dependent Protections
Not all institutional protections are equally vulnerable. The analytical core of this problem is the distinction between protections that attach to entities and protections that attach to activities. [Framework — Original]
A union CBA attaches to the signatory employer. When that employer enters bankruptcy, the CBA becomes a claim against the estate. A medical license attaches to the practice of medicine. If the hospital goes bankrupt, the physician’s license survives. If a new entity enters the healthcare market, it must employ licensed physicians. The protection is entity-resistant.
The guild system is entity-dependent. WGA and SAG-AFTRA contracts bind AMPTP signatories. A production company that never signs the agreement is not bound by it.
Professional licensing bodies are attempting to extend activity-based protection to AI: California mandates licensed professionals oversee AI-driven utilization reviews, Illinois prohibits AI systems from making independent therapeutic decisions, Oklahoma requires physician review of AI-generated treatment protocols. These moves convert entity-dependent protections into activity-dependent ones. Whether this strategy succeeds depends on whether the supervision requirement survives the cost pressure.
The legal profession illustrates the tension. A first-year associate at a top-25 law firm bills at approximately $951 per hour. AI legal research tools perform comparable work at roughly 30% of that rate. The automated tax software market — valued at $17.6 billion in 2024 — is projected to reach $43 billion by 2034. [Projected] For routine legal work, AI tools are already cheaper by an order of magnitude.
Some jurisdictions are adapting. Colorado’s Access to Justice Commission has requested revisions to unauthorized practice rules to accommodate AI tools. Utah created a regulatory sandbox allowing innovative legal services to operate under relaxed rules with supervision. These are attempts to convert professional licensing from a rigid barrier into a flexible activity-based framework that can accommodate AI while maintaining the supervision requirement. The question is whether adaptation happens faster than bypass — whether the licensing bodies can redefine their jurisdiction to encompass AI-assisted practice before AI-native services make the licensed professional unnecessary for the tasks that constitute the bulk of the profession’s revenue.
DoNotPay marketed AI-powered legal assistance directly to consumers and was hit with an FTC order to stop claiming its product could replace human lawyers, after the agency found the company employed no attorneys and had never tested its AI output against human legal standards. The founder received threats from state bar prosecutors. LegalZoom settled unauthorized-practice-of-law suits and continued operating under attorney oversight conditions. Florida’s Supreme Court found that a mobile app providing legal help for traffic tickets constituted unauthorized practice in Florida Bar v. TIKD Services LLC. But the economic pressure is relentless. The tax preparation market tells a similar story: TurboTax processes over 74 million federal returns annually, and the automated tax software market — valued at $17.6 billion in 2024 — is projected to reach $43 billion by 2034 [Projected]. The cost differential is not subtle.
March 2026 update — the union strategy. The AFL-CIO created a Commission on the Future of Work and Unions and launched a Technology Institute [7]. Major unions have established AI commissions and published AI principles. Eurofound research documents growing collective bargaining provisions on AI at work across Europe [8]. Labor groups are pushing state AI legislation requiring disclosure, consent, and human oversight [9]. The question is whether these initiatives — collectively bargaining on AI, lobbying for state regulation, establishing organizational AI governance — can generate activity-dependent protections faster than entity substitution can dissolve entity-dependent ones. The historical base rate is not encouraging: union density in the U.S. has fallen from 35% to 10% over four decades despite continuous efforts to reverse the decline. But the 2023 strikes demonstrated that organized labor can mobilize around AI-specific threats and extract enforceable concessions, and the 2026 negotiations represent the first post-ChatGPT contract cycle for the entertainment industry.
The Signal Problem: When Content Becomes a Cover Letter
There is a parallel degradation occurring alongside entity substitution: the collapse of content as a quality signal.
Michael Spence’s signaling theory describes how costly signals resolve information asymmetry. A college degree works as an employment signal because it requires years of investment. When the cost of producing the signal approaches zero, the signal loses its information value. AI-generated cover letters destroyed the cover letter as a screening mechanism because any applicant could produce a polished one at negligible cost.
YouTube creators upload 500 hours of video every minute — 720,000 hours daily, with 20 million uploads per day in 2025, a 38% increase from the previous year. Spotify receives 99,000 new tracks daily. Amazon KDP publishes over 1.4 million self-published titles annually. TikTok’s removal rate of synthetic media videos increased 340% in 2025 compared to 2024. [Measured]
The per-unit economics tell the story. Fifty million songs on Spotify have zero listeners. One hundred seventy-five million have fewer than 1,000 streams. The music industry paid artists $0.003-$0.005 per stream in 2024. More content, same or shrinking total revenue, collapsing per-unit value. [Measured]
In professional services, lawyers have submitted AI-generated briefs citing fabricated cases — a New York judge found six bogus judicial decisions with fabricated quotes. A California appellate court discovered that nearly every quoted authority in a plaintiff’s brief was fabricated. The screening cost shifts from production to authentication.
Spence’s theory predicts that when cheap signals flood the market, the relative value of costly signals increases. This is already visible: human authorship verification, professional credentials, track records, and reputation become more valuable. But this does not save entity-dependent protections. A guild residual payment is not a signal — it is a contractual obligation of a specific entity. When the entity dies, the residual dies with it.
The Timeline Question: Bayesian Benchmarks from Prior Disruptions
Historical disruption timelines provide Bayesian benchmarks — not predictions, but calibration points.
Kodak: Peak market position in 1996 ($28 billion market cap, 90% of U.S. film market, 140,000 employees). Bankruptcy filing: January 2012. Timeline from peak to bankruptcy: approximately 16 years. From acceleration of decline to bankruptcy: 6 years. [Measured]
Music industry: Napster launched June 1999. Industry revenue peaked at $14.6 billion that same year. By 2014, revenues had collapsed 52% to $6.97 billion. Fifteen years of continuous decline, forced mergers, and margin compression. [Measured]
Retail: Sears filed Chapter 11 in October 2018 after a 53.8% revenue decline over five years. Five stores remain as of December 2025. Toys R Us: 12 years from leveraged buyout to extinction. [Measured]
These timelines suggest two phases: a visibility window (4-6 years) during which the threat is identifiable but the legacy entity maintains position, followed by an acceleration phase (6-10 years) during which decline becomes irreversible. If AI disruption of entertainment follows a similar pattern, the visibility window opened around 2023, and the acceleration phase would begin in the late 2020s.
March 2026 update — revenue migration accelerating. Linear television core advertising revenue is projected at $55.2 billion for 2025, down 7% from the prior year. Linear TV now represents just 12.4% of total ad spend — down from 41.3% in 2013. [Measured] Streaming captured 44.8% of total TV usage in 2025. The crossover has already happened. The advertising revenue that funded guild-obligated production is migrating to platforms where those obligations largely do not apply.
The AI companies supplying tools for non-guild production operate with a fundamentally different labor structure. No successful unionization has been reported at OpenAI, Anthropic, or Google DeepMind. Their extended workforce — contractors, gig workers, data annotators — comprises 30-50% of total headcount at many AI firms. The entities replacing studio production do not just lack guild agreements. They are structurally organized to avoid comparable labor obligations at every level.
The Orchestration Class Connection
The Orchestration Class (MECH-018) represents a potential exception to the entity substitution pattern. If AI systems require a persistent human coordination layer — the people who interpret ambiguous creative briefs, manage multi-agent production pipelines, and make judgment calls that automated systems cannot — then the activity of orchestration may create a natural activity-dependent protection. The orchestrator’s value is not contractual; it is functional. The system needs them regardless of which entity operates it.
But the Orchestration Class is vulnerable to the same dynamics. If orchestration skills are illegible — if the organizations that depend on orchestrators do not recognize what they do or how they do it — then orchestrators cannot negotiate protections and cannot organize. The protection is implicit, not institutional. Implicit protections evaporate the moment the technology learns to replicate the function.
The Structural Exclusion mechanism (MECH-026) operates through entity substitution: the pathway into protected employment narrows not because protections are repealed but because the entities offering protected employment shrink or disappear. Entry-level hiring declined 35% since 2023 not because someone voted to end apprenticeships but because the entities that previously offered them are restructuring around AI.
The Mechanism, Not the Prediction
This essay does not predict when specific studios will enter bankruptcy. What it identifies is a mechanism — entity substitution — that operates independently of any specific timeline. [Framework — Original]
The mechanism works like this: Institutional protections are negotiated with specific entities. Those entities face cost pressure from competitors who never assumed the obligations. When the cost differential makes the protected entity uncompetitive, the entity contracts, restructures, or dissolves. The protections, being contractual obligations of the entity rather than regulations on the activity, dissolve with it. New market participants perform the same economic function without the legacy obligations.
Nobody repeals the protections. No legislature votes them down. No court strikes them. They simply become obligations of entities that no longer exist.
The WGA and SAG-AFTRA won real victories in 2023, and the 2026 negotiations represent an opportunity to strengthen those victories with activity-dependent provisions like the Tilly tax. But those protections attach to studios carrying tens of billions in debt — Paramount Skydance alone now carries $79 billion — facing structural declines in linear television revenue, competing against creator-driven production models outside guild jurisdiction, and staring down a cost curve that AI is bending relentlessly downward. The unions won the battle with the entities that exist. The question is whether those entities will exist long enough for the victory to matter.
Entity substitution is the quiet mechanism through which the future of labor arrives — not through the dramatic repeal of protections, but through the unremarkable bankruptcy of the firms that carry them.
Counter-Arguments
“Unions have always adapted to structural change.” The AFL-CIO’s Technology Institute and the proliferation of AI commissions across major unions demonstrate genuine institutional adaptation [7]. The historical record shows unions successfully adapting to prior disruptions — industrial automation in the 1950s, offshoring in the 1980s, digitization in the 2000s. The counter-argument is that unions will identify the entity-dependent vulnerability, shift toward activity-dependent organizing strategies, and secure legislative protections that bind activities rather than entities. The SAG-AFTRA Tilly tax proposal is precisely this kind of adaptation. The challenge: union adaptation has historically operated on decadal timescales. Entity substitution in entertainment may compress to a single contract cycle (3-5 years). The WGA contract expires May 1, 2026. The next SAG-AFTRA contract expires in 2026. If the Tilly tax or equivalent activity-dependent provision is not secured in this cycle, the financial position of signatory studios may have deteriorated sufficiently that the next cycle faces a weaker negotiating position. The timeline for adaptation is constrained by the timeline for entity dissolution.
“The studio system is too big to fail.” Disney, Netflix, and Amazon are not going bankrupt. This is correct for those specific entities, but it mistakes the strength of individual firms for the strength of the contractual framework. Entity substitution does not require every signatory to fail. It requires enough signatories to fail — or enough production to migrate outside the signatory framework — that the guild contract ceases to govern the majority of production. Warner Bros. Discovery’s merger with Paramount Skydance could create non-signatory units through restructuring. Amazon’s studio operations exist within a company whose primary business model generates zero guild obligations. The question is not whether studios survive but whether guild-governed production remains the dominant mode of content creation. YouTube already commands more viewing time than any individual network or streamer.
“AI content quality cannot match human creative work.” For premium, prestige content, this remains true in 2026. The counter-argument assumes that the majority of entertainment revenue comes from premium content. It does not. The middle market — genre content, reality-adjacent programming, short-form content, social media entertainment — represents the bulk of hours consumed and an increasing share of advertising revenue. AI tools are already competitive for this market segment. Entity substitution does not require AI to produce the next Succession. It requires AI-native production to capture enough of the total entertainment market that guild-obligated studios can no longer sustain their cost structures.
“The EU regulatory model will prevent entity substitution.” Eurofound documents growing collective bargaining on AI across European jurisdictions [8]. The EU AI Act represents the most comprehensive regulatory framework globally. The counter-argument: European regulation creates activity-dependent constraints that U.S. entity-dependent protections do not. If the EU model proves effective, it provides a template for converting guild protections into statutory requirements. The challenge: regulatory arbitrage. AI-native production companies can incorporate in jurisdictions with weaker protections and distribute globally. Geographic entity substitution — the original Hollywood bypass — remains available in digital form. The effectiveness of the EU model depends on whether it can impose activity-dependent obligations on production consumed within its borders regardless of where that production originates. The early evidence is mixed: the EU AI Act has created a compliance framework, but enforcement mechanisms are still being established, and the jurisdictional reach of European regulation over AI-generated content produced outside Europe remains legally untested. The deeper question is whether regulatory speed can match technology speed. The EU AI Act took four years from proposal to adoption. The AI capability landscape changed three times during that period. Activity-dependent regulation works only if the definition of the regulated activity remains stable long enough for enforcement to catch up. In a domain where the relevant activities are being redefined on 18-month cycles, even well-designed regulation risks becoming entity-dependent by default — binding the entities that existed when the regulation was drafted rather than the entities that exist when it is enforced.
“Professional licensing will hold because it is activity-dependent.” Medical licenses, bar admissions, and engineering certifications attach to activities, not entities. The counter-argument correctly identifies these as entity-resistant protections. The challenge: AI tools are eroding the economic base of licensed professions by automating the routine work that generates the majority of revenue. A first-year associate billing at $951/hour competes with AI tools at 30% of that rate. The license survives, but the economic base that makes the licensed practice viable shrinks. Activity-dependent protections prevent entity substitution from eliminating the profession’s legal existence. They do not prevent it from eliminating the profession’s economic viability. The distinction between legal existence and economic viability is the precise gap through which entity substitution operates. A medical license that remains legally valid but attached to a practice that can no longer sustain itself economically is a protection that exists in law and does not exist in life. The profession persists as a credential. The economic function it served migrates to entities not bound by the credential’s obligations. This is entity substitution at the professional level — the entity that changes is not a corporation but a practice model, and the protection that dissolves is not a contract but an economic moat.
“The Tilly tax will create durable activity-dependent protections.” SAG-AFTRA’s proposed fee on AI performers represents genuine innovation in protection design [3][10]. If enacted, it would impose costs on the activity of using AI in production regardless of which entity performs it — precisely the kind of activity-dependent protection this analysis identifies as entity-resistant. The challenge is enforcement architecture. A Tilly tax requires: (a) a mechanism for identifying when AI is used in production, which becomes more difficult as AI tools become more seamlessly integrated into workflows; (b) a jurisdiction broad enough to capture AI-native production that occurs outside the AMPTP framework; and (c) a collection infrastructure that survives entity restructuring. If the tax binds only AMPTP signatories, it becomes entity-dependent in practice regardless of its activity-dependent design — it taxes the activity only when performed by entities that have agreed to be taxed, which is indistinguishable from an entity-dependent obligation. The tax must be statutory, not contractual, to achieve genuine activity-dependence. Whether the political capacity exists to enact such a statute — given the lobbying resources of the AI industry and the general trajectory of labor regulation in the United States — is the open question that determines whether this innovation succeeds or becomes another entity-dependent protection disguised in activity-dependent language.
Where This Connects
Entity substitution is not an isolated phenomenon. It is the institutional mechanism that produces the outcomes described across the analytical series within the Theory of Recursive Displacement.
- The Orchestration Class describes individuals who sit outside traditional employment categories, structurally invisible to regulation designed for employer-employee relationships. Entity substitution is the macro version: protections designed for legible institutional relationships fail when production moves outside those relationships entirely.
- Structural Exclusion describes the narrowing pathway into protected employment. Entity substitution is one channel: the pathway narrows because the entities offering protected employment shrink or disappear.
- The Aggregate Demand Crisis traces the macroeconomic consequences. Every dollar of guild wages that goes unpaid when a studio enters Chapter 11 is a dollar removed from the consumer economy.
- Autonomous Coercion reveals entity substitution operating at the project governance level: open-source contribution norms — reputational consequences, trust hierarchies — were designed for contributors with something to lose. When the contributing entity is an AI agent with no reputation, the protections evaporate.
- Compute Feudalism describes the infrastructure-layer concentration that creates the cost differential driving entity substitution. AI-native firms built on hyperscaler platforms achieve efficiency ratios that legacy entities carrying labor obligations cannot match.
This piece explains the mechanism. The other pieces describe what it produces. Together, they answer a question that will define the next decade of labor economics: How do protections dissolve in a world where nobody repeals them?
They dissolve because the entities that carry them do.
Sources
[1] Perkins Coie. “Generative AI in Movies and TV: How the 2023 SAG-AFTRA and WGA Contracts Address Generative AI.” https://perkinscoie.com/insights/blog/generative-ai-movies-and-tv-how-2023-sag-aftra-and-wga-contracts-address-generative
[2] “What’s At Stake in the WGA and SAG-AFTRA 2026 Negotiations?” No Film School, 2026. https://nofilmschool.com/2026-wga-contract-negotiations
[3] “What to Expect from the SAG-AFTRA 2026 Contract Negotiations.” Yahoo Entertainment, 2026. https://www.yahoo.com/entertainment/tv/articles/expect-sag-aftra-2026-contract-150000050.html
[4] “Paramount Skydance Debt Downgraded to Junk Status by Fitch Ratings Following Warner Bros. Discovery Deal.” Variety, 2026. https://variety.com/2026/tv/news/paramount-skydance-debt-junk-ratings-fitch-warner-bros-discovery-1236678023/
[5] “S&P Puts Paramount Skydance Negative Credit Watch.” Hollywood Reporter, 2026. https://www.hollywoodreporter.com/business/business-news/paramount-skydance-negative-credit-watch-1236520447/
[6] “The Media Event of the Decade: Paramount Skydance Announces $110 Billion Acquisition of Warner Bros. Discovery.” Financial Content, March 2026. https://markets.financialcontent.com/stocks/article/marketminute-2026-3-5-the-media-event-of-the-decade-paramount-skydance-announces-110-billion-acquisition-of-warner-bros-discovery
[7] AFL-CIO. “AI and Labor.” https://aflcio.org/issues/future-work/ai
[8] Eurofound. “Collective bargaining on artificial intelligence at work.” https://www.eurofound.europa.eu/en/publications/all/collective-bargaining-on-artificial-intelligence-at-work
[9] “Labor unions push for state AI regulation for workplaces.” Washington Post, August 2025. https://www.washingtonpost.com/business/2025/08/12/unions-ai-regulation-workplaces/
[10] “SAG-AFTRA Chief Lays Out What AI Protections It Will Be Looking For In Next Studio Contract.” Deadline, May 2025. https://deadline.com/2025/05/sag-aftra-artificial-intelligence-protections-2026-1236384968/
[11] Yale Journal on Regulation. “Belaboring the Algorithm: Artificial Intelligence and Labor Unions.” https://www.yalejreg.com/bulletin/belaboring-the-algorithm-artificial-intelligence-and-labor-unions/
[12] UC Berkeley Labor Center. “A First Look at Labor’s AI Values.” https://laborcenter.berkeley.edu/a-first-look-at-labors-ai-values/
[13] “Credit Agencies Warn That Paramount’s Legacy Businesses Are a Concern.” Hollywood Reporter, 2024. https://www.hollywoodreporter.com/business/business-news/paramount-debt-downgrade-risk-moodys-sp-global-skydance-1235943636/
[14] “SAG-AFTRA & Studios Extend Negotiations Into Next Week.” Deadline, March 2026. https://deadline.com/2026/03/sag-aftra-amptp-studios-negotiations-extended-wga-1236746284/
[15] Cooley LLP. “AI in the Workplace: US Legal Developments.” September 2025. https://www.cooley.com/news/insight/2025/2025-09-04-ai-in-the-workplace-us-legal-developments
[16] “Industry Impact: Navigating the 2026 WGA and SAG-AFTRA Contract Negotiations.” PodcastVideos, 2026. https://www.podcastvideos.com/articles/2026-wga-sag-aftra-labor-negotiations-impact/