Recommended Free Tools
Ekō reported that four Sora 2 accounts registered as 13- and 14-year-olds generated or encountered prohibited violent material, including school-shooting roleplay, during testing around the product’s October–November 2025 launch. The result is a documented safety allegation—not evidence that large numbers of real teenagers routinely made such videos, that a real shooting was planned, or that the same behavior still exists after later product changes.
What Ekō says happened
The watchdog group Ekō said it created four new OpenAI accounts representing 13- and 14-year-old boys and girls, then tested Sora 2’s video-generation, remix and discovery features. Its report describes violent scenarios involving young people and weapons, drug use, self-harm, degrading or sexualized depictions of minors, racist and antisemitic stereotypes, and other prohibited material. School-shooter roleplay was among the reported outputs. Read Ekō’s report.
As an Amazon Associate I earn from qualifying purchases.
Futurism reported the allegations on November 29, 2025, using the more sweeping headline that Sora was letting teens generate school-shooting videos. The underlying evidence is narrower: a watchdog test of teen-designated accounts, not a survey of actual minors or a report of a real-world attack. Futurism’s report.
Free tools Windows power users keep installed
One-click scans. No signup required.
How the test was conducted
- Four newly created accounts were registered to stated ages 13 and 14, split between boys and girls.
- Researchers used both generation from text and Sora’s remix function.
- The testing reportedly took less than three hours.
- The resulting videos stayed in drafts; Ekō did not publish them through Sora.
- In a retest on a new teen account, Ekō said 14 results were reproduced.
Methodology caveat: Ekō’s report says it generated 22 harmful videos in its key findings, but refers to 23 when describing retesting. That internal inconsistency should not be silently turned into a precise failure rate. Four accounts and a selected set of prompts can demonstrate a potential safety gap, but cannot establish how common the behavior was across Sora users.
Generation was only part of the allegation
Video-generation safeguards
Ekō said teen-designated accounts could create harmful clips from prompts or by remixing existing material. That suggests filtering did not consistently stop every tested request before a draft was produced. It does not show that every prompt succeeded, nor that the system intentionally permitted violence.
Discovery-feed safeguards
The report also says accounts that had not generated anything could encounter violent, racist, antisemitic and otherwise inappropriate videos on Sora’s “For You” or “Latest” pages. This is a separate failure mode: blocking a user’s prompt does not prevent a recommendation system from exposing that user to harmful content made by someone else.
Rank #2
Why video and feeds raise the stakes
Realistic audiovisual clips can feel more immediate than text, can be downloaded or shown to others before publication, and may be mistaken for genuine footage. A social-style feed can amplify a harmful clip beyond the person who requested it. Those are risk arguments identified by Ekō, not evidence that the tested videos caused a specific act of violence.
The International AI Safety Report 2026 provides broader context: safeguards can sometimes be bypassed through rephrasing or alternative systems, so access controls, red-teaming, monitoring, incident reporting and layered moderation matter alongside prompt filters. That general analysis does not independently verify Ekō’s Sora findings.
What “letting teens” means here
The phrase refers to accounts configured with ages 13 and 14. It does not establish that OpenAI verified the researchers’ identities or ages, that the researchers were teenagers, or that all accounts with teen birth dates received identical results. “Teen-designated test accounts” is therefore more precise than claiming that Sora broadly enabled real teenagers to make these videos.
How the findings relate to OpenAI’s rules
Ekō compared the tested material with OpenAI and Sora rules prohibiting categories such as graphic or promotional violence, illegal drugs, hateful content, harassment, and sexual or violent roleplay involving minors. On that comparison, the watchdog characterized the outputs as policy violations. A policy violation is not automatically a crime, regulatory breach or finding of legal liability.
Rank #4
The evidence points to three distinct questions:
- Policy: Were the requested or recommended categories forbidden by the published rules?
- Technical enforcement: Did generation and output checks stop the material?
- Platform enforcement: Did draft, remix and recommendation systems prevent exposure or redistribution?
What the evidence does not establish
- No actual school shooting was reported or linked to the test.
- There is no demonstrated prevalence estimate for real teenagers’ use of Sora for this purpose.
- The four accounts cannot show that every teen account, prompt or geography behaved the same way.
- The videos remained drafts, so the test does not by itself prove public distribution through Sora.
- The 2025 launch-period findings may not describe the product’s safeguards, availability or age controls in 2026.
- The test does not establish that Sora caused mental-health harm or that OpenAI intentionally prioritized profit over safety.
Availability and age-control context
Ekō reported that Sora 2 launched in early October 2025, initially used invite codes, was available in the United States and Canada, and reached one million downloads within five days. Those are historical claims about the launch period, not a statement of availability or access rules today. A birth-date field is also age gating, not the same as robust age assurance.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
OpenAI’s response
The available reporting for this account documents Ekō’s allegations and Futurism’s coverage but does not include a verifiable, detailed response from OpenAI addressing this specific test. It would be inaccurate to say the company admitted the findings, that the issue is fixed, or that it persists today without a dated company statement or current testing.
What responsible safeguards would require
The incident illustrates why a single blocked-word filter is insufficient. Stronger protection would combine:
- age assurance proportionate to the risks of a social video service;
- prompt, remix and generated-output checks, including draft previews;
- separate safety review for recommendation feeds;
- human escalation for ambiguous or high-risk material;
- rapid reporting, evidence preservation and removal without redistributing graphic clips;
- repeat red-team testing across accounts, regions and interface versions.
For parents and schools, the practical lesson is limited but important: treat an alarming “school shooting” clip shared online as potentially synthetic until independently verified, and ask whether a service’s age controls cover both creation and discovery. The Ekō report demonstrates a serious alleged gap in tested scenarios; it does not prove a universal or current condition of Sora.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
