On September 30, 2025, OpenAI launched a new app called Sora. In a stark transition for the company, Sora is in many ways similar to traditional social media, rather than using a chatbot format like OpenAI’s ubiquitous ChatGPT. Much like Instagram or TikTok, Sora consists of short-form videos, usually 10 to 30 seconds long, that users can create and post. The difference, of course, is that every Sora video is AI-generated.
At face value, the application is harmless enough. Users can type in prompts to create their own videos, including the use of others’ likenesses if they give permission. This content is then made into a feed much like other social media apps. Beneath the surface, however, there are a variety of issues.
The first and most obvious problem is deepfakes. Deepfakes are AI videos that are created in order to seem real to mislead viewers. Generally, they depict people doing or saying things they did not do or say. In one 2019 example, Facebook founder Mark Zuckerberg was depicted talking about harvesting and controlling data from billions of people. Deepfakes have existed for years, but Sora increases both the ease of creation of the videos and the frequency with which people will see them. To combat this, OpenAI requires a watermark (a symbol attached to a video to identify it) on Sora content to distinguish it as AI. Unfortunately, according to Time Magazine, tools have already been created to add these symbols to real videos and remove them from AI ones. This has resulted in deepfakes making their way into Sora and other social media platforms. According to NPR, OpenAI’s other security mechanism is a system that refuses prompts it deems to be a misuse of the system. However, in a short time frame using Sora NPR was able to get around the guardrails, creating videos depicting President Nixon stating that the moon landing was fake (a recurring conspiracy) and Niel Armstrong taking his helmet off on the moon. Considering this required only a small amount of effort from a few NPR staffers, the wider threat of deepfakes is clear.
Attempts have been made to avoid this problem. However, legislation against AI content has proven to be difficult to create. In August 2025, a federal judge struck down a California deepfake law, which aimed to restrict AI-generated deepfake content during elections, after Elon Musk’s X sued the state on the basis that the law violated First Amendment protections. This has created an environment in which AI content is frequently difficult to discern from its human-created counterparts.
The general impact on social media and its users is likely to be widespread as well. Given Sora’s propensity to create fake videos that are too realistic to be easily known as such, the trustworthiness of social media content is likely to be impacted as well. This is concerning because the use of AI content is likely to accelerate the polarization effect of social media. Even before Sora, in 2023 the Niskanen Center reported that social media was known to play a role in driving political polarization. This occurs largely through what the Center refers to as “echo chambers”, in which social media algorithms feed users content from only one side of an issue until they are driven to one side or the other of the political spectrum. Consider that this polarization occurred using real videos, whether misleading or not. Now, totally false content can be created. Political enemies can be demonized for actions they never took and things they never said, with a real-looking video to back it up. It may become difficult to discern the truth of any political issue through the AI mess. The impact on polarization, therefore, could be more notable than anticipated.
One final problem is the fact that Sora could have a negative impact on social media in general. As AI content floods traditional social media, it may become difficult to discern what is and is not real. This could result in people giving up social media, believing false information, and anything in between. What is certain is that the effects of Sora AI will be widespread, serious, and are coming very, very soon.
