Skip to content

Topic

AI training data licensing and consent

Legal and business question of whether AI developers must license or pay rights holders to train models on copyrighted text, code, images, or audio.

Current stories

investConfirmed5 publishers

Meta's new business terms appear to let it scrape publisher sites for AI training from October 30

Meta's commercial terms, effective October 30, appear to let it scrape sites tied to business accounts for AI training, according to Nieman Lab. With no opt-out, a publisher negotiating payment for its archive is bargaining with a counterparty that may already hold a contractual route to the same text.

Perspective Coverage

5 publishers
Builder
Builder 17%
Operator
Operator 51%
Investor
Investor 32%

Reality

Evidence58
Adoption
Not scored
Hype gap+30
Incentives50
Confidence62
buildOne report1 publisher

Atmosfy's Atlas repackages its consumer video map of 100 million places as licensed AI data

Atmosfy launched Atlas on October 7 to license what it says are more than 500 million place-tagged video experiences to AI, search and commerce builders. The counts are the company's own and unaudited, so buyers will need to check recency, place matching and rights coverage on any slice they license.

Publishers:runtimewire.com

Reality

Evidence30
Adoption
Not scored
Hype gap+40
Incentives70
Confidence40
productConfirmed4 publishers

Suno moves every user onto v6 because the Warner settlement required it

Three new models, one of them free, and no route back to the ones the labels sued over. The retirement was a settlement term rather than a product call, which tells you who the roadmap answers to.

Perspective Coverage

4 publishers
Builder
Builder 44%
Operator
Operator 26%
Investor
Investor 30%

Reality

Evidence62
Adoption
Not scored
Hype gap+30
Incentives72
Confidence60