Natasha Lyonne says AI has an ethics problem because right now it’s ‘super kosher copacetic to rob freely under the auspices of acceleration’

Natasha Lyonne’s Asteria Film Co., co-founded in 2022, has partnered with Moonvalley AI to launch Marey, a generative AI video tool that uses only open-license or explicitly permitted source material; the service is available publicly via a credits-based subscription starting at $14.99 per month. Asteria’s approach targets legal and reputational risks that have dogged other AI video models by prioritizing copyright-safe inputs, signaling a potential niche standard for studio-grade generative content even as Lyonne stresses human oversight and limited current use in actual film/TV production.

Analysis

Market structure: Permissioned-AI studios (Asteria, Moonvalley-type businesses) and licensing platforms (e.g., Shutterstock) are the direct beneficiaries as they can monetize clean datasets and command higher per-title fees; large infrastructure providers (Google/Alphabet, OpenAI-type models) face pricing and reputational pressure that can compress margins. Expect a 12–36 month shift where content owners gain bargaining power, raising the marginal cost of compliant training data 20–50% relative to scraped alternatives and slowing free-model proliferation.

Risk assessment: Tail risks include adverse judicial rulings or national-level copyright laws that could impose damages >$500M–$1B on major model providers and effectively ban certain training practices; these are highest-probability within 3–18 months as litigation and legislation accelerate. Hidden dependencies include concentration of rights-holders and compute providers; a choke-point here would spike model costs and force consolidation. Key catalysts: blockbuster lawsuits, high-profile studio-platform deals, and next-gen model launches within 30–180 days.

Trade implications: Near-term (weeks–months) prefer asymmetric, hedged exposure: long media/licensing specialists and modest protective shorts or put spreads on dominant model hosts (GOOGL/GOOG). Rotate 3–5% of equity weight from pure infra into media/IP licensing over 1–2 quarters. Options: use 3–9 month verticals to limit downside while keeping upside optionality.

Contrarian angle: The market underestimates the commercial upside of “permissioned AI” — historically (music publishing post-Napster) licensing regimes created durable cashflows; downside is fragmentation and higher content costs that favor mid-cap licensors. If regulatory pressure forces compliance, expect accelerated consolidation and 30–60% re-rating opportunities for compliant IP owners over 12–24 months.

Key Decisions for Investors

• Establish a 2–3% long position in NFLX over 6–12 months: buy in two tranches (60% now, 40% on a 5–10% pullback) to play studios benefiting from permissioned-AI distribution and increased content monetization.
• Implement a 0.5–1% short/hedge against GOOGL/GOOG via a 3–6 month bear put spread (buy 10% OTM put, sell 5% OTM put) sized to limit max loss to ~0.25% portfolio while capturing regulatory/IP downside risk.
• Add 1–2% long position in Shutterstock (SSTK) as a licensing beneficiary; deploy 50% now, 50% on an 8–12% pullback, horizon 3–9 months to capture contract re-negotiations and platform demand.
• Allocate 0.5–1% notional to options: buy a 6–9 month NFLX call spread 25–35% OTM (cap upside cost) and a 3–6 month GOOGL put spread 10–15% OTM to keep downside protected while limiting premium outlay.
• Monitor 30–90 day catalysts (major copyright lawsuit verdicts, federal/state AI legislation, and any studio–AI vendor deals); if an adverse legal ruling occurs, reduce NFLX/SSTK exposure by 50% and increase GOOGL downside protection to 2% notional.