Data rights and licensing

Settled rules on what models may train on and who gets paid.

Facts

Lane: Data
Wave: 6 of 12 (build order, not a date)
Needs: 0 boxes all the way back through hard links
Unlocks: 8 boxes downstream through hard links
Leverage: 1.8 points · stands in the way of 6 of 53 stages · rank 5 of 57
Cluster: Around licensed generative media

Evidence as of October 2026

A US court found training on lawfully obtained books can be fair use, but not on pirated copies. Anthropic agreed to pay at least $1.5bn, about $3,000 a book.

01Anthropic settles landmark AI copyright lawsuit for at least $1.5bn

Around it

Needs · 0

Nothing beyond the detail list. It can start at any time.

Unlocks · 3

Paying the people who make knowledgecapability · Speculative
Licensed generative mediastage · Emerging
Training data supplycapability · Built today · helps, not required

Loops and off-ramps

Synthetic content and the commonsharmful loop
Training data ruled outoff-ramp · hit directly

Openings researchers' judgement, October 2026 scouting

Startup room 75/100 · rank 5 of 124 Research room 75/100 · rank 3 of 124

Crowding4 Lab overlap4 Small-team fit2 Buyer today4 Research openness4

Depth: deep
Players: RSL Collective (RSL 1.0); CLA and CCC collective licences; Cloudflare 402; labs via settlements (Anthropic $1.5bn)
Wedge: Training-corpus acquisition-provenance audit (proving no pirated copies) and licence-registry tech for collective bodies; research on payment-grade training-data attribution. startup