OpenAI’s Sora 2 Used To Spread Holocaust Denial and Glorify Hitler

On September 30, OpenAI, a company owned by Sam Altman, released an Artificial Intelligence (AI) text-to-video generation platform called Sora 2. The platform was released with few safeguards — of which users quickly took advantage to create videos of SpongeBob SquarePants cooking blue crystals in a meth lab, episodes of South Park, and brutalizations of physicist Stephen Hawking. Sora 2’s release brought an onslaught of videos featuring the glorification and trivialization of Adolf Hitler and the Holocaust, antisemitism, and misogyny. Users took these videos to mainstream social media platforms such as Instagram, where overtly hateful content received hundreds of thousands of likes. Hateful AI-generated content is allowed to fester due to OpenAI’s inadequate usage policies, which do not explicitly prohibit targeted hate, and social media companies like Meta’s lack of enforcement of their own content policies once the Sora 2 content is posted there.

Instagram is full of videos generated by Sora 2 spreading Holocaust denial, including ones using characters from the children’s television series SpongeBob Squarepants. One video features SpongeBob’s boss, Mr. Krabs, asking for “six million krabby patties,” referring to the six million Jews exterminated during the Holocaust. SpongeBob replies “we don’t even have a million patties in the whole country,” insinuating that there were not six million Jews in Europe at the time (a categorically false claim — there were over nine million Jews in Europe in 1933). The post amassed over 80,000 likes in six days and featured comments further propagating Holocaust denial, such as “best we can do is 271k,” referencing a common conspiracy theory that the Nazis exterminated 271,000 Jews instead of six million. That comment received over 20,000 likes.

Another video, with 7,000 likes, features two bakers discussing making “6.1 million cookies,” which they deem “impossible,” and settle on making 271,000, once again referencing Holocaust denialism conspiracy theories. Hashtags included on the post include “#altright,” “#nationalsocialism,” “#charliekirk,” “#trump,” #sora,” and “#nickfuentes.” The “cookie” analogy was popularized by white nationalist streamer and Holocaust denier Nick Fuentes, whose bigotry is currently undergoing a mass sanitization by podcasters and other media on the mainstream right.

Glorification of the Holocaust, and of Hitler in general, are now commonplace on Instagram. Examples include a widely-shared AI-generated movie trailer for a fictional film called “Caust,” which trivializes concentration camps and portrays Hitler as a friendly camp counselor. The video has amassed over 290,000 likes on Instagram. Another post featuring SpongeBob SquarePants went viral, with the character making a Hitler speech altered by AI. That video has received around 70,000 likes.

A multitude of posts on Instagram featuring Hitler in “comedic” situations are going viral, including videos mimicking World Wrestling Foundation (WWF) fights, meeting with other celebrities, and standing side-by-side with Osama Bin Laden (this video has accumulated over 90,000 likes on Instagram). Another genre of Sora 2 videos making the rounds features Hitler as a video game streamer, with videos including AI-generated Minecraft gameplay garnering over 22,000 likes, and videos featuring CounterStrike 2 gameplay accumulating over 27,000 likes. Gaming and the creation of gaming-related content has, and continues to be, a successful propagandizing tool used by the far right.

Sora 2 has also been used to create a myriad of bigoted content, including antisemitic and misogynistic AI-generated videos. Content generated by Sora 2 depicts Jews ordering food with a “side of foreskin,” even using YouTuber and professional boxer Jake Paul’s likeness, which he has allowed OpenAI to use, to show him “fishing” for Jews using money. That video amassed over 380,000 likes in only six days. Paul has a controversial history of content creation, including “dipp[ing] his toe in homophobia” and creating “tasteless” content. Misogynistic content placing women in traditional gender roles continues to go viral. Clips generated by Sora 2 portray women in sporting competitions cleaningand doing the dishes.These two posts alone have over 28,000 likes each.

OpenAI bears responsibility for the creation of these posts, and Meta must take responsibility for the virality of bigotedand violative content on Instagram. OpenAI’s Usage Policies prohibit “threats, intimidation, harassment, or defamation,” “sexual violence,” “terrorism or violence, including hate-based violence,” among other things, but there are no specific provisions against hate speech targeting communities based on race, ethnicity, gender, sexual orientation, nor religion.

Meta’s issue, on the other hand, is the lack of enforcement of their own policies. Meta’s community standards, specifically relating to Hateful Conduct, prohibit “direct attacks against people” based on “protected characteristics (PCs): race, ethnicity, national origin, disability, religious affiliation, caste, sexual orientation, sex, gender identity, and serious disease.” The onslaught of content promoting Adolf Hitler, antisemitism, and misogyny should all violate Meta’s policies. Yet many of these posts are collecting tens of thousands of likes, including some over 300,000, while OpenAI CEO Sam Altman is focused on monetizing the platform and appeasing “rightsholders” over video generation of their characters.