AI Copyright Fight Tests What Counts as Publisher Permission

artificial intelligence copyright microsoft openai paid content paywalls publishing Sep 21, 2026

Publishers have spent years deciding what readers can see for free and what sits behind a paywall. AI is creating a different question:

What are AI companies allowed to do with that content?

Newly public court filings in the copyright fight involving The New York Times, other publishers, OpenAI and Microsoft show just how unsettled the answer is. The publishers allege that OpenAI and Microsoft acquired and used copyrighted journalism without permission, including content behind paywalls or subject to other restrictions. The filings include an allegation that OpenAI employees discussed a way to get around the Times' paywall.

OpenAI and Microsoft dispute the publishers' claims and argue that major parts of their use are protected by fair use.

Here's the part publishers should notice: OpenAI and Microsoft are also arguing that, for some uses, publishers had ways to say no and didn't use them.

That puts a practical question in front of publishers: What actually counts as saying no?

A paywall may not answer every AI question

A paywall controls what a subscriber sees, but it may not control how every automated system accesses or uses the underlying content. That is part of what makes these cases interesting.

The publishers argue that OpenAI and Microsoft acquired articles from behind paywalls or contrary to other restrictions, then used that journalism in commercial AI products.

OpenAI makes a separate argument about ChatGPT's Browse feature, which retrieves current information from the web. It says publishers knew how to block its crawler through robots.txt or a hard block and argues that copies made through Browse before publishers opted out were implicitly licensed.

Microsoft makes a similar argument about Copilot's use of current web content. It says it introduced a NOARCHIVE control in 2023 that allowed publishers to remain in traditional Bing search while preventing their pages from being used to help Copilot answer a user's question. Microsoft argues that publishers who knew about the control and did not use it implicitly permitted that use.

Those are arguments OpenAI and Microsoft are asking the court to accept. The court has not ruled that failing to use an AI-specific control automatically gives an AI company permission.

OpenAI and Microsoft are also making separate fair-use arguments about AI training. That is a different question from whether an AI system had permission to retrieve current content from a publisher's site.

Search and AI use are becoming different decisions

Most publishers don't want to disappear from search. A publisher may want Google or Bing to index an article so someone searching for the topic can find it and visit the site. That does not necessarily mean the publisher wants the same article used to train an AI model or provide the substance of an AI-generated answer.

This is not only playing out in court. The technology is changing too.

On September 15, Cloudflare introduced a "Disallow AI Training" setting designed to let website owners continue participating in traditional search while separately declining AI training by crawlers that support the setting.

It is not one universal switch. Cloudflare says Apple and Google support the approach. Microsoft is still developing support for a site-level no-training signal for Bing, targeted for early 2027. Publishers using Bing today may need Microsoft's existing controls, including NOARCHIVE, depending on what they want to restrict.

Cloudflare is also separating automated access by purpose, including traditional search, AI training and agents acting on behalf of users. For publishers, that means more control, but also more decisions.

A paywall manages the subscriber experience. Authentication can protect subscriber-only areas. Terms of Use describe permitted behavior. Robots.txt, meta tags and other technical settings can send instructions to automated systems. Those pieces may not all say the same thing.

More publishers are taking the issue to court

This fight is moving beyond the largest national publishers. On September 16, another 26 U.S. publisher organizations representing roughly 160 local, regional and specialty publications filed a federal lawsuit against Microsoft and OpenAI.

The group includes Times Publishing Company, publisher of the Tampa Bay Times, The Austin Chronicle, Florida Trend and a range of local, nonprofit and specialty publishers. Their complaint also alleges that OpenAI and Microsoft copied content from behind paywalls and other access restrictions.

Those allegations have not been decided by a court.

Insider Take

Publishers already know that AI companies want valuable content. The harder question is whether they know what their own technology is currently telling those companies they can do with it.

For years, a publisher could think about the paywall primarily from the subscriber's side: Who gets access? What requires registration? What requires payment?

AI adds another layer. Search visibility, AI training, AI-generated answers and agent access can now be different decisions, and the controls can differ from one platform to another.

The law is still catching up, but the court fight shows why publishers should understand their current setup now. Publishers are pointing to paywalls and access restrictions as evidence that their content was not freely available for any use. OpenAI and Microsoft are pointing to crawler controls and opt-outs and arguing that publishers had ways to say no more specifically.

So there is a basic question worth taking back to your own team:

What uses of our paid content are we allowing today, and where have we actually said yes or no?

For publishers, associations, research companies and other businesses selling access to proprietary content, that is becoming part of managing the product subscribers pay for.

Related Member Resources

The legal fight is still developing, while the technology and access controls are changing too. These resources put the court activity in context and connect it to the broader AI decisions subscription operators are now making.

Enforcement, Litigation & Risk Activity Tracker
Follow significant lawsuits, enforcement actions, settlements and court decisions affecting subscription businesses. The Tracker makes clear where each matter stands, which is especially important when allegations like those in the OpenAI and Microsoft cases are still being tested in court.
AI for Subscription Operators
Plain-English AI updates, operator lessons and practical guidance for subscription businesses. The collection follows how AI is changing areas such as content, customer experience and operations as those changes become real business decisions.