New York Times reporter files lawsuit against AI companies

Investigative reporter John Carreyrou and five other authors have filed suit against xAI, Anthropic, Google, OpenAI, Meta and Perplexity alleging their large language models were trained on copyrighted books without permission, with the complaint intentionally brought by individuals rather than as a class action. The filing highlights Anthropic’s recent $1.5 billion settlement with authors and contends class members in that case recover only a tiny fraction (about 2% of the Copyright Act’s $150,000 statutory ceiling); Perplexity has stated it does not index books and the case is the first to name xAI as a defendant.

Analysis

Market structure: Plaintiffs (NYT, individual authors, legacy publishers like DIS) are net beneficiaries — licensing revenue could rise by tens to low‑hundreds of millions if industry-wide deals follow the Disney‑OpenAI precedent; technology incumbents (GOOGL/GOOG, META, AAPL exposure to LLMs) face higher marginal costs and potential profit margin compression of ~1–3% of operating income if licensing becomes standard. Competitive dynamics favor firms with proprietary/enterprise data and deep pockets to absorb settlements; smaller AI entrants and open‑source models face either extinction or need for sponsored licensing. Cross-asset: expect near‑term option IV lift in big‑cap AI names (+20–50% relative), modest widening of IG tech credit spreads (10–30bps) if litigation escalates, and temporary USD strength during risk‑off flows.

Risk assessment: Tail risks include injunctive relief forcing model retraining (weeks) or statutory damages up to $150k per infringed work producing aggregate exposures in the low‑billions for large crawls — a low‑probability, high‑impact scenario within 6–24 months. Immediate (days) risk is IV/price volatility; short term (weeks–months) is settlements/licensing that reprice business models; long term (years) is tighter IP regimes that permanently increase cost of LLM training. Hidden dependencies: undisclosed third‑party data brokers, indemnity gaps in enterprise contracts, and insurance policies that may not cover punitive statutory damages. Catalysts: judge rulings, Anthropic settlement terms, and new licensing agreements (Disney/OpenAI style) will accelerate outcomes.

Trade implications: Tactical positions should be asymmetric — hedge downside in tech while selectively long content owners. Consider small protective allocations: buy 3‑month 10% OTM puts on GOOGL and META to cover 1% portfolio delta, and establish 1–2% long positions in NYT and DIS with 6–12 month horizons to capture licensing upside. Pair trade idea: long DIS (1.5% weight) vs short GOOGL (1% weight) for 6–12 months; corporate license wins for DIS should re‑rate legacy media multiples. Sector rotation: trim high‑beta AI ad names into media/IP owners and cloud/enterprise AI (favor providers with contractual data control).

Contrarian angles: The market may be overstating existential risk — big AI vendors can amortize settlements (a $1.5B bill is <0.2% of a $750B cap tech company) and negotiate broad licenses, so dramatic multiple compression is unlikely for highest‑quality franchises. Historical parallel: music streaming litigation led to structural royalty costs but ultimately stable TAM growth and normalized multiples within 2–4 years. Unintended consequence: litigation could accelerate paid, permissioned data markets (new recurring revenue for publishers) and increase moat for cloud incumbents that broker licensed datasets.

Key Decisions for Investors

• Establish a 2–3% long position in NYT (NYT) over 6–12 months to capture potential licensing revenue upside; set a 15% stop‑loss and reassess if no licensing headline in 90 days.
• Purchase 3‑month puts at ~10% OTM on GOOGL and META sized to protect 1% of portfolio exposure each (buy puts covering 1% notional) to hedge elevated IV and litigation risk; if an adverse injunction/certification occurs, increase hedge to 2% within 5 trading days.
• Implement a pair trade: go long DIS (1.5% portfolio weight) via a 6‑month 20%/40% call spread to limit capital at risk, and short 1% notional of GOOGL stock to express relative upside of content licensors versus AI ad/infra names.
• Reduce gross exposure to high‑beta AI ad names by 2–4% and redeploy into cloud/service companies with contractual data control (favor names with enterprise contracts) over the next 30–90 days; trigger reallocation back to tech if a major settlement reduces implied damages by >50% within 90 days.