Lawyers behind $1.5 billion Anthropic settlement slash fee bid after pushback from judge

Lead: Plaintiffs' counsel cut their attorney-fee request to $187.5M (12.5% of the $1.5B Anthropic copyright settlement) from an initial $300M bid. Judges and Anthropic objected to fees being allocated to three non-appointed firms, prompting the reduction; the filing says the $187.5M request is based solely on class counsel work. The settlement requires Anthropic to pay class members more than $3,000 per copyrighted work and destroy certified pirated datasets; final approval is scheduled for an April 23 hearing.

Analysis

This settlement sequence creates a durable commercial pathway for authors and rights-holders to extract rent from AI model training — the obvious effect is more licensing demand, but the second-order outcome is emergence of specialized compliance and rights-clearance vendors who can scale marginably cheaper than bespoke settlements. Expect a flurry of bilateral licensing talks over the next 6-18 months as platform-scale AI vendors seek predictable unit economics for training data rather than litigate.

Near-term legal optics matter: the April 23 final-approval hearing is a binary catalyst that will either crystallize a template for class-wide recovery or open an appeals window that prolongs uncertainty for months; either outcome changes negotiating leverage materially for both publishers and AI firms. If the judge trims fees further or orders reallocation, plaintiffs’ firms may pursue alternative compensation mechanisms (e.g., contingency carve-outs paid outside the class fund), which would create messy follow-on litigation and delay monetization of the settlement for content owners.

From an industry-structure perspective, large, diversified technology incumbents with balance-sheet scale and existing licensing relationships (enterprise software, search, cloud providers) are best positioned to absorb recurring content-licensing costs; smaller pure-play AI vendors face margin compression or must pivot to verticals with proprietary data. For investors, the dominant trade is to reweight toward companies that sell licensing, rights-management and legal/analytics services to publishers and corporates, and to underweight small-cap model vendors lacking scale or defensible distribution.

Key Decisions for Investors

• Long RELX (RELX.L) or RELX ADR-equivalent, 6–18 month horizon: buy for exposure to rights-management, licensing workflows and corporate legal analytics. Target +20–30% upside if RELX captures incremental content-licensing services; downside -12% if enterprise budgets reallocate away from paid content.
• Long TRI (Thomson Reuters), 3–12 months: buy into demand for authoritative legal records, settlements tracking and IP-clearance tools. Risk/reward ~2:1 with modest downside if ad-revenue softness persists; take profits post-April 23 hearing.
• Pair trade: long MSFT (or GOOGL) / short small-cap AI training plays (select names with <$500M market cap that monetize via generic LLM inference), 12–24 months: incumbents can internalize licensing costs and up-sell cloud AI services while small vendors suffer margin squeeze. Expect 15–25% relative outperformance for the long leg under a precedent-setting licensing regime; monitor earnings for margin pressure in smaller names as stop-loss.
• Event hedge around April 23: buy a low-cost put spread on a basket of mid/small-cap AI vendors (30–90 day expiries) to protect against a wave of adverse rulings or appeals that re-introduce legal overhangs. Cap cost at small premium for asymmetric protection against a 10–30% drawdown in the cohort.