Skip to content
Submit incident
AII-1002-58237
Well documentedThree or more independent publishers report the core facts.

Sora 2 generated fake children's toy ad referencing Jeffrey Epstein

OpenAI's Sora 2 video generator was used to create a mock children's toy commercial set on Jeffrey Epstein's island, featuring references to sex trafficking and a Trump caricature. The video, created intentionally by AI filmmaker Mathieu Samson as apparent satire to highlight safety gaps, spread widely online and drew criticism for trivializing child exploitation and violating the system's content policies.

October 2025 Β· Mathieu Samson Β· Harmful and abusive content
October 2025
AII-1002-582374 sources ↓
Quick facts
Category
Harmful and abusive content
Country
United States
Source
nytimes.com
Date
October 2025
Drafted by
Raidu Incident Detection

What happened

Mathieu Samson, founder of AI filmmaking firm Kickflix, created a video ad using OpenAI's Sora 2 that mimicked a 1990s children's toy commercial set on Jeffrey Epstein's private island.

The video featured palm trees, secret rooms, spy cameras, and an "Orange Man" action figure resembling Donald Trump saying "don't release the files," alluding to sex trafficking and abuse scandals linked to Epstein.

Reported impact

The video spread rapidly on X and TikTok, triggering backlash about trivializing child exploitation and violating Sora 2's rules against depicting public figures.

Critics raised ethical and social concerns about the content referencing child exploitation, while others questioned the opportunistic nature of its creation for publicity.

Index analysis

OpenAI's safety documentation states Sora 2 filters block unsafe, sexual, and other potentially damaging depictions of public figures, but the Epstein ad violated these safeguards.

Samson acknowledged insufficient filters, telling The Washington Post, "There definitely is not enough filters right now," and predicted users would continue finding workarounds despite restrictions.

Governance controls

OpenAI stated that Sora 2 has filters designed to block unsafe, sexual, terrorist, self-harm, and disparaging depictions of public figures, but the video's creation indicated inadequacies in these guardrails.

Classification

Category
Harmful and abusive content
Industry
Arts, entertainment and recreation
Country
United States
Harm status
Potential harm
Intent
Intentional
Timing
Post-deployment
Harm types
Psychological, Public interest
Harmed
Children, General public
Failure modes
Deepfake video, Content moderation failure
AI systems
Sora 2

Sources and evidence

This record was researched and written by the Index, based on the following original reporting.

nytimes.com
Sora 2 generated fake children's toy ad referencing Jeffrey Epstein
theguardian.com
Sora 2 generated fake children's toy ad referencing Jeffrey Epstein
futurism.com
Sora 2 generated fake children's toy ad referencing Jeffrey Epstein
newsweek.com
Sora 2 generated fake children's toy ad referencing Jeffrey Epstein