OpenAI’s ChatGPT, developer-facing API, and the newly launched Sora experienced a service disruption on Wednesday, leaving users unable to access the platforms for several hours.
According to OpenAI’s status page, the outage began around 11pm WAT (3pm PT) and continued until approximately 5am WAT (9pm PT) when services were mostly restored.
In a brief statement, OpenAI acknowledged the incident on X (formerly Twitter), stating, “ChatGPT, API, and Sora were down today but we’ve recovered.”
The announcement drew a cryptic response from Elon Musk, who replied, “@Grok 😉,” indirectly inviting users to leverage his new AI platform.
OpenAI’s status updates revealed that users are gradually regaining access to the affected platforms.
During the outage, visitors to ChatGPT’s website encountered an error message stating that the issue had been identified and was being resolved.
Adding to the day’s complications, OpenAI launched its integration with Apple’s iOS 18.2, which some speculated might be connected to the outage.
However, Edwin Arbus, OpenAI’s developer community lead, clarified on X, “The OpenAI outage was unrelated to 12 Days of OpenAI or Apple Intelligence. We made a config change that caused many servers to become unavailable.”
Sora’s Launch by OpenAI Overwhelms Servers
The outage coincided with the public debut of Sora, OpenAI’s video-generation AI tool, which had already faced capacity issues earlier in the week.
CEO Sam Altman acknowledged the overwhelming interest in Sora, admitting the company had underestimated the demand.
Users reported difficulty generating videos due to server overload on launch day.
Sora’s functions, which include generating short videos based on text or image prompts, have attracted huge interest, particularly among gaming and creative communities.
However, its training data, which reportedly includes publicly available sources such as video game streams, has raised questions about intellectual property rights and potential copyright violations.
The training of generative AI models like Sora on unlicensed content has increasingly become a contentious issue.
Experts say that if video game playthroughs and related materials are part of Sora’s training data, OpenAI could face legal challenges.
Copyright attorneys highlight the layered intellectual property rights inherent in-game content, including game developers’ copyrights, streamers’ unique contributions, and user-generated content.
“Training an AI model on unlicensed footage from games risks infringing on multiple layers of copyright protection,” noted Joshua Weigensberg, an IP attorney.
This issue could bring about legal challenges, as courts are trying to find a balance between fair use and creators’ rights.
Nonetheless, OpenAI continues its “12 Days of OpenAI” event, a product launch series leading up to the holiday season.
Recent announcements have included updates to its o1 reasoning model, enhancements to the Canvas platform, and the integration with Apple Intelligence.
While the company remains tight-lipped about Sora’s training specifics, it maintains that its tools adhere to industry standards.
However, with generative AI becoming more integrated into creative industries, OpenAI and similar companies may face increased investigations over the ethical and legal dimensions of their operations.
