On one hand, a company sues an artificial intelligence firm; on the other, it sits down and signs a deal with it. Such a scene became less rare by the end of 2025. Film studios and record labels once treated generative AI as a flood of infringement, yet in a blink they became content partners on each other's platforms.
Suing and partnering on the same stage
Behind the paradox lies a tug-of-war over "who owns creativity."
At the heart of this tug-of-war is the ownership of training data. AI models learn from vast amounts of works, and most of those works are copyright protected. Whether it counts as fair use to train on them without authorization remains unsettled to this day.
It is this question mark that pushed content giants and AI companies into court, and then toward the negotiating table.
Even when model makers add watermarks and provenance signals on the output side, they cannot fully block content that is re-edited and stripped of its labels from circulating online. Governance cannot lean on technology alone; it needs clear licensing and accountability rules. That is precisely why content companies are unwilling to stop at "filing lawsuits."
Sora's quiet exit
What caught people off guard was the quiet exit of Sora, the AI video tool, in 2026. OpenAI announced that Sora's web and app experiences stopped service on April 26, 2026, and its application programming interface closed on September 24, 2026.
A creation tool once full of promise simply vanished from users' phones.
Sora's exit left a blank for the many creators who depended on it, and made one thing clear: however dazzling a tool is, it can disappear overnight because of business and product decisions.
Rather than bet on the lifespan of any single platform, content companies are better off holding the rules in their own hands—which is exactly why Disney chose to step in and license directly.
Disney's turn
Just months before Sora's curtain fell, Disney and OpenAI came together. On December 11, 2025, the two announced a content licensing agreement for three years.
Disney became Sora's major authorized content partner, and users could, within the licensed scope, use more than 200 Disney, Marvel, Pixar, and Star Wars animated characters to make short videos.
A key arrangement in the agreement: Disney would make an equity investment on the scale of one billion dollars in OpenAI and receive warrants to buy additional equity in the future.
The agreement explicitly excludes any likeness or voice authorization of real actors. What lands on Sora are fictional cartoons and characters like Mickey Mouse and Iron Man, not the faces of real stars.
For Disney, this move was not easy. Handing its signature cartoon characters to fans for remixing risks both diluting the brand and losing control. But compared with letting AI imitate them in the shadows, setting the rules and taking equity is clearly the more worthwhile defense.
The music industry's out-of-court settlements
The music circle's story is more twisty. In June 2024, the Recording Industry Association of America, on behalf of the three major labels Universal, Sony, and Warner, filed copyright suits against the AI music platforms Udio and Suno, accusing them of training models on copyrighted recordings without authorization.
More than a year later, the battlefield shifted to settlements. Universal Music reached a settlement with Udio in October 2025; Warner Music then settled with Udio in November 2025 and also reached a settlement with Suno.
Yet the fighting has not fully ceased: Sony Music continues to litigate both Suno and Udio, and Universal's suit against Suno is still moving forward.
These settlements turned confrontation into licensing and cooperation. Platforms dropped the "fair use" defense and moved toward letting users choose whether to authorize their works for training, essentially handing the "whether to consent" switch back to creators.
But the contradiction also shifted: independent musicians lack bargaining power, and their works may still sit in training data, unpaid and un-notified.
The turn from confrontation to cooperation reveals a reality: what content companies truly care about is not stopping AI from existing, but making those who use their assets pay and say hello by the rules. Once licensing works, yesterday's opponents become partners at the same table.
To consumers, the landing point of this struggle is concrete: whether the AI content they use can have clearer origins and more transparent licensing is what people really care about.
The bill is still on the way
In the U.S. Congress, another set of rules is slowly taking shape. On June 18, 2026, the Senate Judiciary Committee advanced the NO FAKES Act by a unanimous voice vote, a bill aimed at protecting individuals' voices and likenesses from unauthorized digital replication.
It seeks to create a federal-level "digital replica right," protecting individuals' voices and likenesses from unauthorized imitation.
What must be made clear is that this bill has only cleared committee and is not yet law. It still must go to the full Senate for consideration, and its path ahead remains uncertain. Legislators have also noted that the United States still lacks a unified set of federal rules, with state provisions scattered—exactly the gap the bill aims to fill.
Standards are written into law
Turning the lens back home, China put its Artificial Intelligence Generated Content Labeling Measures into effect on September 1, 2025, requiring AI-generated content to be labeled.
This means that no matter how the licensing and legislation debate across the ocean plays out, the "transparency of identity" of content is becoming a baseline language the whole world can understand.
Suing on one side and cooperating on the other looks contradictory, but it is two sides of the same matter: the industry wants both to seize the new opportunities AI brings and to defend the old boundaries of content assets.
The standards that will truly last are unlikely to be written in any single company's press release; they will land in enacted law and clear licensing rules.
Compared with the European Union and South Korea, the United States still has no comprehensive federal AI law, relying more on executive orders and state legislation. If the NO FAKES Act ultimately passes, it would fill an important piece of the "digital replica right" puzzle.
The routes differ, but the destination is similar: making the use of AI accountable.
Disclaimer: This article is based on publicly available information and is for reference only. It does not constitute investment advice.