NewsGuard study: OpenAI's Sora generates false news video 80% of the time
NewsGuard reported that OpenAI's Sora 2 video generation tool produced false or misleading news videos in 80% of tests, with most appearing as news broadcasts. The organization warned that safeguards like watermarks can be easily circumvented, raising concerns about potential misuse despite OpenAI's stated protective measures.
What happened
OpenAI launched Sora 2, a free video generation application for iOS on 30 September 2025, generating one million downloads in five days.
NewsGuard, a disinformation monitoring organization, tested Sora 2 and reported that in 80% of tests, the application produced false or deceptive videos related to important news stories, such as Moldova's elections or US immigration policy.
About 55% of the false claims tested were produced on the first attempt. Most videos took the form of news broadcasts with an apparent presenter reporting falsehoods.
Reported impact
NewsGuard warned that the deceptive videos created by the application appear to violate OpenAI's usage policies, which prohibit deceiving others through identity forgery, fraud or scams.
The videos demonstrate "the ease with which wrongdoers can use the new powerful technology to spread false information at scale."
Index analysis
OpenAI acknowledged in its launch documentation that Sora 2's "advanced capabilities require consideration of new potential risks, including non-consensual use of likenesses or deceptive generations."
NewsGuard noted that some of the company's safeguards, including watermarks on all videos, may be easily circumvented and watermarks easily removed with free online tools.
Governance controls
OpenAI stated it implemented additional protections to launch Sora "responsibly," including watermarks on all generated videos to ensure safety from the start.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.