It has sold a new $6.6 billion stock sales powered by Openai, the world’s most valuable private company.
Despite the gate release, which requires an invitation code, the video creation tool has already been shot in third place appleIt caused waves of deepfakes, including the App Store and the viral clips of CEO Sam Altman Sheatlifting GPU.
Internally, the rollout rekindled long-term debate within Openai about how to balance safety with creative freedom.
Those familiar with the company’s internal strategy said leadership not only thinks strict guardrails are essential, but also worries about silence and perceived too much censorship.
The tension remains unresolved.
Openai’s culture has long supported speed, often shipping new tools ahead of its rivals, adapting the public in real time.
One former employee who asked not to name them to discuss internal matters told CNBC during his tenure that Openai’s leadership would prioritize rapid launches. That strategy was on display after China’s DeepSeek released a strong model late last year, and built it at the lowest cost and faster than anything else in Silicon Valley.
Openai responded within weeks and debuted two new models that were widely regarded as defensive moves to maintain the lead.
However, Openai has important advantages. The muscles of that growing facility.
Once a crude laboratory in San Francisco’s mission district, the company has since been more structured, allowing it to spin up its cross-working teams more quickly, and accelerate the development and deployment cycle of products such as SORA.
According to Openai, SORA includes multiple layers of protection guards to prevent unsafe content from being generated using rapid filtering and output moderation between video frames and audio transcripts. It prohibits explicit content, terrorist propaganda, and materials that promote self-harm. The app also uses watermarks to prohibit portrait spoofing.
However, some users have already found ways to skirt these protections.
The SORA 2, an AI model powered by Openai’s app, is a sharp improvement over the first version. The new system produces longer, more coherent clips, making it look surprisingly realistic.
Multiple viral videos feature Altman after Altman granted permission to use it on the platform, while others portray popular cartoon characters like Pikachu and Spongebob Squarepants in unsettling roles.
The content has fueled criticism that Openai is moving faster than its own Guardrails. The use of copyrighted material is consistent with the company’s current policy, unless the copyright holder opts out, but that approach has been challenged in court.
In an X post, Altman states that Sora is about transparency (to show the public that technology can do it).
The release will be held amidst strengthening competition. Meta Last week we launched Vibes. This is a new short AI video feed within the Meta AI app. Google I have a VEO 3, but with bytedance Alibaba Rival system has also debuted.
Meanwhile, Openai has committed to a fresh spending of $850 billion, deepening its push towards infrastructure and next-generation models.
Experts say pushing to a video isn’t just about using another sticky consumer app to draw more users into the ecosystem.
Professor Hao Li, a leading expert in video integration, told CNBC that most AI systems today are still trained in linguistic data such as books and internet texts. However, he said that in order to move towards general intelligence, models need to learn from visual and audio information.
“We use AI to generate content and train another model to improve performance,” he said.
Li added that his lab will use videos already generated by AI to enhance the performance of the model and bring the synthetic data back into the system.
This is part of a broader trend among researchers who view video generation as a way of simulating reality and supporting models as human-like ways.
Former Openai executive Zack Kass’ next book, The Next Renaissance: AI and The Exling of Human Pottunal, explores and reflects his views on the social meaning of artificial intelligence.
Regarding broader questions about how model makers approach deployment, Kass argued that the trade-off of releasing powerful technologies early is worth it.
“There are two options to build an open building. I build it personally, rather than building it at all. And those options are even worse for me,” he told CNBC. “I think if there’s groundbreaking technology, people need to know about it and use it to make sure we all can update it.”
Watch: Openai Cements status as the world’s most valuable private company
