Sources: OpenAI offered some media companies as little as between $1M and $5M per year to license their news articles for use in model training, a tiny amount
OpenAI has offered some media firms as little as between $1 million and $5 million annually to license their news articles for use …
Context & Ripple Effects
The offer puts a price signal on negotiations that had already been underway for months with major publishers. It follows a dispute in which Dow Jones said OpenAI had no agreement to use Wall Street Journal reporting, underscoring that licensing was becoming a live commercial and rights-management issue rather than a theoretical one.
OpenAI had also paired content access with funding and developer credits in its American Journalism Project arrangement, while subsequent talks with CNN and Fox pointed to a broader push for text, video, and image rights. The small reported cash range makes the economic terms—not just whether deals exist—the key point of comparison.
First-order effects
- Media companies receiving these proposals must decide whether the reported $1M–$5M annual payment justifies granting training rights to their archives.
- OpenAI gains a concrete negotiating position in publisher talks already reported to include major news groups, setting an initial benchmark for news-content licenses.
Second-order effects
- Publishers with differentiated archives or broader multimedia rights can press for better terms, as negotiations with CNN and Fox over text, video, and image licensing suggest that rights bundles may matter as much as a single headline payment.
- A low opening price increases the incentive for publishers to compare direct licensing with alternatives such as partnerships, credits, or withholding content pending clearer terms.
Third-order effects
- If such agreements proliferate, news archives may become a tiered input market: broad, nonexclusive training access could command different prices from premium or multimedia rights.
- The pattern points toward more formalized content-rights negotiations around AI training, but the durable pricing model will depend on whether publishers can establish leverage beyond one-off bilateral deals.
The trend: Generative-AI developers are moving from informal web-data dependence toward negotiated commercialization of content rights, with the value of each corpus still being tested.