Gizmodo reports new standard to charge AI for web content training use

The emergence of a technical standard to charge for AI training on web content is another sign that the era of “free data” is closing. For a decade, large models have quietly feasted on the open web; now publishers are pushing back with treaty‑like mechanisms encoded in standards and protocols. If widely adopted, such a standard could significantly change the cost structure of frontier-model training: instead of treating the web as a commons, labs would need to negotiate access or face clearer legal and reputational risk.

For the race to AGI, the impact cuts both ways. On one hand, higher and more structured data costs could slow down brute-force scaling and encourage more efficient use of existing corpora, plus greater reliance on synthetic and user-generated data. That might delay some capabilities but improve accountability and signal alignment with societal expectations. On the other hand, large, well‑capitalized players are best positioned to absorb licensing costs and join standard-setting processes, which could entrench incumbents and make it harder for new labs to compete. Over time, we may see a bifurcation: a licensed, high‑quality data layer used by a few giants, and a more constrained or synthetic layer for everyone else.