The wild, copyright-free playground that OpenAI briefly created with its powerful new video-generation model, Sora 2, has officially come to an end—after lasting barely a week.
When Sora 2 first launched, OpenAI adopted a copyright “opt-out” policy, meaning that copyright holders had to actively request that their characters or works not be used in Sora-generated videos. But following a flood of controversy and growing pressure from major entertainment companies, CEO Sam Altman announced a dramatic reversal: the company is now shifting to an “opt-in” model.
Under this new policy, rightsholders will need to explicitly grant permission for their characters or content to appear in AI-generated videos. Altman explained that this approach will “give rightsholders more granular control over generation of characters.”
However, this move hasn’t gone down smoothly among Sora’s growing fanbase. The users who made the platform explode in popularity—by creating and sharing viral clips using familiar, copyrighted characters—are furious.
Sora’s Creative Chaos Meets Corporate Reality
The backlash might have been inevitable. From the moment Sora launched, its accompanying TikTok-style social app became flooded with AI-generated videos featuring well-known, copyrighted characters. Users gleefully created surreal and often controversial mashups—everything from Disney icons and anime heroes to, bizarrely, a version of SpongeBob in Nazi uniform.
Unsurprisingly, that didn’t sit well with the copyright owners. On Monday, the Motion Picture Association (MPA) stepped in, publicly urging OpenAI to halt the potential wave of copyright infringement happening on its platform. Within hours, OpenAI responded and adjusted its policy.
In a blog post, Altman framed the shift as a win-win, saying that OpenAI was listening to creators and studios alike. “We’re hearing from a lot of rightsholders who are excited about this new kind of ‘interactive fan fiction,’” he wrote, “but want the ability to specify how their characters can be used (including not at all).”
Despite the optimistic tone, Altman acknowledged that the system isn’t foolproof. “There may be some edge cases of generations that get through that shouldn’t,” he admitted. “Getting our stack to work well will take some iteration.”
Copyright Holders Aren’t Entirely Convinced
The entertainment industry, however, doesn’t seem reassured. MPA CEO Charles Rivkin fired back, saying that OpenAI must take full responsibility for preventing infringement—not push that burden onto rightsholders. “Well-established copyright law safeguards the rights of creators and applies here,” Rivkin said, emphasizing that OpenAI must ensure compliance across its AI products.
And while OpenAI is now giving creators control over outputs, critics point out that the inputs—the data used to train Sora—remain murky.
A report from The Washington Post suggested that the first version of Sora was likely trained on a vast amount of copyrighted material without permission. OpenAI has not confirmed whether it obtained proper licenses for Sora 2’s training data, but the model’s uncanny ability to recreate copyrighted characters with near-perfect accuracy has raised eyebrows across the industry.
The Bigger Copyright Battle in AI
The situation echoes a larger legal and ethical debate playing out across the AI industry. Earlier this year, Anthropic paid $1.5 billion to settle a lawsuit filed by authors who accused the company of using their books without consent to train its models. Although one court ruled that using copyrighted material for AI training may qualify as “fair use,” that decision remains controversial—and far from universal.
OpenAI, meanwhile, has lobbied the Trump administration to officially classify AI training as fair use, a move that could shield companies from liability. For now, it appears that OpenAI is testing boundaries, balancing innovation with legal risk, and hoping that forming the right alliances will prevent future fallout.
Users Are Furious — and Vocal About It
While the policy change may have appeased Hollywood, it’s angered the people who made Sora viral. According to 404 Media, online communities like Reddit and X (Twitter) are now overflowing with frustrated users.
Many Sora fans say the ability to play with copyrighted characters was what made the tool so entertaining. One Reddit user vented, “That was the only reason this app was fun.” Another accused OpenAI of caving to “moral policing and leftist ideology,” claiming it’s “destroying America’s AI industry.”
In short, the Sora community is not handling the change well. What began as a freewheeling creative experiment has now become a battleground between user freedom and corporate control.
A Lesson in Creative Limits
OpenAI’s move highlights a growing tension at the heart of the AI boom: how far can creative tools go before crossing legal lines?
Sora 2 proved that people are eager to use AI to remix and reimagine the cultural icons they love—but it also proved that copyright law still holds immense power, even in the age of generative media.
For now, OpenAI is trying to walk a tightrope—keeping studios happy without alienating its fanbase. Whether this new “opt-in” system strikes the right balance remains to be seen. But one thing’s clear: the era of AI’s copyright-free chaos didn’t last long. And for many users, that feels like the end of the fun.