Scarlett Johansson accuses tech companies of ‘theft’ to train AI

A coalition of high-profile creators led by the Human Artistry Campaign’s Stealing Isn’t Innovation movement — including Scarlett Johansson, Cate Blanchett, Vince Gilligan and members of R.E.M. — has published an open letter accusing major tech firms of using copyrighted works without authorization to train AI models and demanding ethical licensing partnerships. The group highlights reputational and legal risk for AI platforms and points to existing licensing alternatives, while Johansson cited a recent dispute over an OpenAI voice resembling her that prompted a pause by the company. For investors, the push raises the prospect of increased litigation, licensing costs and regulatory scrutiny for AI firms that could affect AI training practices and margins over time.

Analysis

Market structure: Copyright pushback benefits incumbent content owners (DIS, SONY) and marketplaces that can broker licensing; expect 5–15% incremental revenue opportunity for large studios/labels over 12–24 months if licensing markets form. Tech incumbents (MSFT, GOOGL, META) face higher marginal costs for training data and greater legal/operational friction, but core cloud/compute vendors (NVDA, AMZN AWS) remain structurally insulated as compute demand is sticky. Pricing power shifts toward vertically integrated players that own both models and licensed content; small AI startups reliant on scraped data will see negative unit economics and higher customer acquisition costs.

Risk assessment: Tail risks include a landmark class action or injunction (probability ~10–20% over 12 months) forcing retroactive licensing payments >$500M for a major model operator, or new US/EU rules requiring provenance and opt-in—both would compress operating margins 200–500bps for model operators. Short-term (days–weeks) expect reputational volatility and elevated IV in AI names; medium-term (3–12 months) legal filings and settlements; long-term (1–3 years) structural licensing market and compliance costs reshape TAM. Hidden dependencies include indemnities from cloud providers, distribution contracts, and private-equity-backed AI plays that lack balance-sheet depth to absorb settlements.

Trade implications: Tactical allocations: overweight content owners (DIS, SONY) and NVDA compute exposure; underweight/trim small-cap pure-play AI/data firms (AI, SOUN) and venture-backed model providers. Use options to size downside: buy 60–120 day puts on concentrated AI incumbents if a regulatory filing >$500M appears, and buy 6–12 month call spreads on NVDA to capture continued GPU demand. Sector rotation toward Media/Entertainment and Infrastructure over next 3–12 months is preferred; reduce allocation to enterprise AI software startups by 30–50%.

Contrarian angles: Consensus frames this as an existential threat to AI; underappreciated is monetization — a forced licensing market could create durable, annuity-like income for studios worth 10–25% of current content segment profits over 2 years. Reaction is likely underdone for content-owner equities and overdone for small-cap AI names priced for unrestricted data access. Historical parallel: music piracy litigation (2000s) led to licensing platforms and durable revenue for rights holders; similar outcome could uplift DIS/SONY, while pruning speculative AI entrants.

Key Decisions for Investors

• Establish a 3% long position in Disney (DIS) and a 2% long position in Sony (SONY) within the next 30–90 days to capture potential licensing revenue; target a 12–24 month horizon and an upside target of +10–25% vs current levels if licensing deals materialize.
• Reduce exposure to small-cap pure-play AI/data companies (e.g., C3.ai (AI), SoundHound (SOUN)) by 40–60% within 30 days; these firms face disproportionate legal and capital risk if scraping becomes costly—cut if market cap < $3bn and gross margins < 40%.
• Buy a 6–12 month NVDA call spread (e.g., pick a notional equal to 1–2% portfolio) to play secular GPU demand; simultaneously buy 60–120 day ATM puts equal to 0.5–1% notional on MSFT or GOOGL as insurance against a regulatory shock exceeding $500M in alleged damages.
• If a major lawsuit or congressional action is announced seeking >$500M or an injunction within 60 days, trim mega-cap AI equity exposure (MSFT, GOOGL, META) by 5% absolute allocation within 5 trading days and redeploy proceeds into DIS/SONY or short small-cap AI names.
• Monitor 30–60 day catalysts: (1) any class-action filings against OpenAI/MSFT/GOOGL, (2) US/UK/EU legislative hearings on AI copyrights, and (3) public licensing agreements announced by large studios; act (enter/exit) within 5 trading days of these triggers.