OpenAI has announced a significant change to the copyright policy for its new AI video generation tool, Sora, following a wave of user-generated videos featuring unauthorized use of popular characters. CEO Sam Altman acknowledged the company was unprepared for the strong reaction from the public and stakeholders, prompting a shift from an opt-out system to one that gives intellectual property owners direct control.
Key Takeaways
- OpenAI is changing its Sora copyright policy, requiring rightsholders to grant permission for their characters to be used, a reversal from its initial opt-out approach.
- The decision came after users created numerous videos depicting copyrighted characters like SpongeBob SquarePants and Pikachu in controversial situations.
- CEO Sam Altman admitted the company misjudged how people would react to AI-generated video, stating it felt "more different to images than people expected."
- Future updates to Sora will include more robust controls for intellectual property owners to manage how their content is used on the platform.
Initial Launch Sparks Copyright Controversy
When OpenAI launched its text-to-video application, Sora, it implemented a copyright policy that placed the burden on media companies and creators. The initial framework was an opt-out system, meaning all copyrighted characters were usable by default unless the rightsholder specifically requested their removal.
This approach led to immediate and widespread use of well-known intellectual property. Users of the app, which functions with an endless scroll similar to TikTok, began generating 10-second videos featuring famous characters in unexpected and often problematic scenarios. Reports surfaced of videos depicting characters like SpongeBob SquarePants in Nazi-themed content and Pikachu in criminal acts.
The rapid proliferation of these unauthorized videos created a significant public relations challenge for OpenAI and raised immediate questions about the platform's content moderation and respect for intellectual property rights.
Understanding Sora's Functionality
Sora is an AI tool designed to generate short videos from simple text prompts. Users can describe a scene, and the AI will create a 10-second video clip, complete with audio. The platform also allows individuals to create an AI-generated version of themselves, called a "cameo," provided they have given consent for their likeness to be used.
OpenAI Announces Major Policy Reversal
In response to the growing controversy, OpenAI CEO Sam Altman announced a complete reversal of the policy. The company is abandoning the opt-out model in favor of a system that empowers copyright holders directly.
"We obviously really care about rightsholders and about people," Altman stated, confirming the company's new direction. He announced that OpenAI would now "let rightsholders decide how to proceed."
This change effectively shifts the platform to an opt-in model. Under the new rules, intellectual property cannot be used on Sora unless the owner has explicitly granted permission. This move is intended to prevent the unauthorized creation of content featuring protected characters and give creators and media companies full control over their assets.
Altman added that while many rightsholders are interested in the technology, they require "a lot more controls" before they are comfortable participating. The company is now focused on building these necessary tools.
Altman Admits Company Underestimated Reaction
During a discussion about the policy change, Sam Altman provided insight into the company's thinking, admitting that OpenAI had been surprised by the intensity of the public and stakeholder response.
"I think the theory of what it was going to feel like to people, and then actually seeing the thing, people had different responses," Altman explained. He suggested that the tangible nature of AI-generated video had a different psychological impact than static AI images.
Video vs. Image: A Different Impact
Altman's comment that video "felt more different to images than people expected" points to a growing area of study in human-computer interaction. The addition of motion and sound can make AI-generated content feel more real and potentially more threatening when used to depict beloved characters in negative contexts.
The CEO also noted that the platform's rapid growth contributed to the problem. "Sora got very popular very quickly," he said. "We thought we could slow down the ramp; that didn’t happen." This surge in user activity amplified the copyright issue, forcing the company to react much faster than anticipated.
Future Controls and the Path Forward
OpenAI's focus is now on developing a more sophisticated set of controls for the Sora platform. These tools will be designed to give rightsholders granular management over their intellectual property, potentially allowing them to set specific restrictions on how their characters can be depicted.
Altman expressed optimism about the future, suggesting that major content owners will eventually make their properties available on Sora once the appropriate safeguards are in place.
"We want to build these additional controls, but I think you will see a lot of major pieces of content be available with restrictions about what they can do and not do," he concluded. The incident serves as a key learning moment for OpenAI as it navigates the complex ethical and legal landscape of generative AI, particularly in the realm of video content where issues of likeness and copyright are especially sensitive.