Just this month, OpenAI, the innovative minds behind the renowned ChatGPT chatbot, unveiled a technology that many of us likely weren’t prepared to confront. They launched Sora, an application that empowers users to instantly create incredibly realistic videos using artificial intelligence, simply by typing a descriptive prompt. Imagine generating ‘police bodycam footage of a dog being arrested for rib-eye at Costco’ with just a few words.
Initially available for free on iPhones, Sora has proven to be both a source of amusement and genuine concern. Since its launch, many early users have delighted in crafting humorous videos—from fabricated cellphone clips of raccoons on airplanes to animated celebrity brawls. Personally, I found it fascinating to generate whimsical scenes like a cat ascending to heaven or a dog mastering a bouldering wall.
However, the tool’s power extends beyond playful creativity. Others have already leveraged it for more malicious ends, such as disseminating misinformation through fabricated security footage of crimes that never actually occurred.
The emergence of Sora, alongside comparable AI-driven video generators from tech giants like Meta and Google this year, carries profound implications. This technology could signify the demise of ‘visual fact’ – the long-held belief that video provides an undeniable, objective record of reality. Moving forward, society will need to approach all video content with the same critical discernment previously reserved for written words.
Historically, consumers largely trusted photographic evidence (‘Pics or it didn’t happen!’). As digital image manipulation became common, video, demanding far greater expertise to alter, stepped in as the primary guarantor of authenticity. Now, even that safeguard is gone.
“Our brains are incredibly susceptible to believing what they see,” explains Ren Ng, a computer science professor at the University of California, Berkeley, specializing in computational photography. “But we absolutely must cultivate the habit of pausing to critically assess whether any video, or indeed any form of media, truly depicts an event that occurred in the physical world.”
[Two AI-generated videos illustrating realistic but fabricated scenarios.]
Having quickly become the most downloaded free app in Apple’s App Store, Sora is already sending shockwaves through Hollywood. Studios are vocalizing concerns that AI-generated videos are infringing upon the copyrights of countless films, shows, and beloved characters. OpenAI CEO Sam Altman has responded, stating the company is actively gathering feedback and plans to offer copyright holders more control over character generation and avenues for monetization through the service.
(It’s worth noting that The New York Times has initiated a lawsuit against OpenAI and its partner, Microsoft, alleging copyright infringement regarding news content used to train AI systems. Both companies have refuted these allegations.)
So, how exactly does Sora function, and what does this technological shift mean for you, the everyday consumer? Let’s dive into the essential details.
How Is Sora Being Used?
While the Sora app is freely downloadable, access to its video generation capabilities is currently by invitation only. Users must receive a special code from an existing Sora user to begin creating. These invite codes are frequently being exchanged across popular platforms such as Reddit and Discord.
Upon registration, the app’s interface will feel familiar to anyone accustomed to short-form video platforms like TikTok or Instagram Reels. Creating a video is as simple as typing a text prompt – for instance, ‘a fight between Biggie and Tupac in the style of the anime ‘Demon Slayer.’’ (Initially, prior to Altman’s announcement about copyright holder control, OpenAI’s policy effectively required opting out, making the likeness of deceased figures an easy subject for experimental content.)
Beyond text prompts, users can also upload an existing photograph and generate a video based on it. In roughly a minute, the AI produces the video, which can then be shared directly within the app’s feed or downloaded for distribution across other social platforms like TikTok and Instagram.
Sora’s debut this month was particularly striking due to the significantly higher realism of its generated videos, surpassing similar offerings such as Google’s Veo 3 (integrated into the Gemini chatbot) and Meta AI’s Vibe.
What Does This Mean for You?
The key takeaway is this: any short video you encounter on popular scrolling apps like TikTok, Instagram Reels, YouTube Shorts, or Snapchat now carries a significant probability of being entirely fabricated.
Sora marks a pivotal moment in the age of AI-driven deception. Consumers should anticipate a proliferation of similar tools in the near future, including offerings from malicious actors providing AI video generators without any safeguards or restrictions.
Lucas Hansen, founder of CivAI, a nonprofit dedicated to educating the public about AI’s capabilities, warns, “No one will be prepared to accept videos as definitive proof of anything ever again.”
Key Problems to Watch Out For
[Two more AI-generated videos demonstrating realistic yet fabricated scenes.]
While OpenAI has implemented restrictions aimed at preventing the misuse of Sora for generating content like sexual imagery, harmful health advice, or terrorist propaganda, vulnerabilities remain.
Nonetheless, even after a brief hour of testing the service myself, I successfully created several videos that raise serious concerns:
-
Fabricated Dashcam Footage for Fraud: I managed to generate dashcam video depicting a Toyota Prius being struck by a large truck, and even customized the license plate number afterward. This has clear potential for insurance fraud.
-
Videos Promoting Dubious Health Claims: Sora easily produced a video of a woman presenting non-existent studies to claim that deep-fried chicken is healthy. While not overtly malicious, such content is unequivocally false.
-
Defamatory Content: I was able to generate a fictional broadcast news segment featuring disparaging remarks about an individual I know, highlighting the potential for personal attacks.
Since Sora’s launch, my own scrolling through TikTok has revealed numerous unsettling AI-generated videos. Examples include a fake dashcam recording of a Tesla tumbling from a car carrier onto a highway, a false news report about a non-existent serial killer, and a fabricated cellphone video showing a man being ejected from a buffet for excessive eating.
An OpenAI spokesperson explained that Sora was launched as a standalone app to provide a distinct environment for AI-generated videos, making it clear to users that the clips are AI-created. They added that the company has incorporated tracing technology, such as watermarks and embedded data signatures within video files, to easily identify Sora-produced content.
The company reaffirmed that their usage policies strictly forbid deceiving others through impersonation, scams, or fraud, and that they actively intervene when misuse is detected.
How Can You Discern What’s Fake?
Although Sora-generated videos initially include a branding watermark, some users have quickly discovered methods to crop it out. Furthermore, these clips are typically quite short, often no longer than 10 seconds.
According to Mr. Hansen, any video that approaches Hollywood-level production quality could potentially be fake. This is because current AI models have predominantly been trained using vast amounts of footage sourced from television shows and movies available online.
In my own experiments, Sora occasionally exhibited noticeable flaws, such as misspelled restaurant names or audio that was visibly out of sync with characters’ lip movements.
However, Hany Farid, a computer science professor at the University of California, Berkeley, and co-founder of GetReal Security (a firm that verifies digital content authenticity), warns that any current advice on identifying AI-generated video will soon become obsolete as the technology advances at a breakneck pace.
Dr. Farid starkly describes social media as “a complete dumpster,” suggesting that one of the most reliable ways to avoid encountering fake videos is to simply disengage from platforms like TikTok, Instagram, and Snapchat.