By Sara Guaglione  •  September 23, 2026  •

Ivy Liu

“I think where we end up is very unclear,” he said. “It’s very hard to get these things going. It takes years, and so a marketplace that is valuing content in real time for use in AI training or inference is just going to take a long time.”

AI content licensing deals are increasingly difficult to structure because publishers can’t predict how their content will be used, or what today’s rapidly evolving AI platforms will look like several years from now, Greenberg said.

There are still a lot of questions that need to be answered, he said, such as: “How is content used? What is the valuation? How do you prevent it from being a race to the bottom on price downstream? Where does the content go? Who owns the outputs? … And then of course, is the revenue enough? Is it worth it if you’re a publisher to break apart your content and sell it for fractions of a cent or whatever the price is?”

But partnerships with AI companies and platforms can give publishers some sway to push for improved linking, attribution and referral mechanisms to help drive users back to their sites, Greenberg said.

“There is a way to nudge these things forward. And being in a partnership is a good way to do that. You certainly have more influence if you’re engaged with someone than if you’re not,” he said.

The New York Times’ partnerships team has doubled to 10 people since Greenberg joined last summer, after working in partnership roles at Google for web and search initiatives aimed at publishers and creators. The Times’ partnerships team works closely with product, audience, communications, legal and executive teams when evaluating AI licensing deals, he said.

Greenberg outlined the core principles The New York Times sticks to when evaluating those potential deals, and why it’s proven to be “tricky” to figure out the conditions in AI licensing marketplaces.

The New York Times’ three core principles for evaluating potential AI licensing deals include:

• The right value exchange: in order for AI companies and tech platforms to use the Times’ content, there needs to be an agreement in place, or express permission, per Greenberg.
• Sustainability of the partnership: “It can’t just be this one-time lump sum, and then there’s no engagement, there’s no partnership beyond revenue. We want to engage with partners, think of new ways to collaborate. We think a lot about driving the ecosystem forward and doing things that will benefit other publishers and other content creators,” Greenberg said.
• Control and guardrails: The Times needs to have control over how content is used, and where it’s used by the AI platforms. “That can be tricky, but that’s important to us,” Greenberg noted.

The Times won’t engage in any deal that doesn’t have any of these principles, according to Greenberg. Ultimately, The New York Times is a subscription business, and so the main goal is to “build a direct relationship with an audience and to build daily habits,” Greenberg said. The Times won’t engage in any deal that “would get in the way of building that relationship,” he said.

The New York Times’ AI licensing partnership with Amazon checks off those boxes: it is a fair value exchange, a collaborative partnership, and there are “the right sort of guardrails and conditions that allow us to be comfortable entering into it,” Greenberg said.

Greenberg declined to share if the Times has any other existing AI licensing partnerships.

“The reality is that there are not that many deals that have been signed in the last few years… because those conditions are tough to meet in a lot of different ways,” he said. “Increasingly, we hear all of the AI labs and all of the platforms say, ‘We believe in the sustainability of content creation, we want content creators to be able to make the revenue they need to make. We respect IP and copyright.’ They say all of these things, but then also the devil is really in the details.”