"It's been almost three weeks since OpenAI launched Sora 2, the latest AI tool that lets anyone generate and share videos of just about anything - with only limited restrictions. The potential for trouble was obvious from the start. Within hours, users created racist depictions of Martin Luther King Jr., videos of Pikachu being barbecued and SpongeBob posing as Adolf Hitler."
"I can't speak specifically to what fact-checkers are seeing. But I will say, based on a scrolling experiment I've done every day on Instagram since Sora came out, about 25% to 30% of my feed is Sora videos that have been crossposted as Reels. The very first day it came out I got sent three separate ones in earnest. The watermark isn't super noticeable when you're scrolling quickly. It happened lightning fast compared to Midjourney coming out and the swagged-out Pope fooling everyone."
OpenAI launched Sora 2 nearly three weeks ago, enabling broad generation and sharing of AI-created videos with few restrictions. Within hours, users produced racist depictions of Martin Luther King Jr., a barbecued Pikachu and a depiction of SpongeBob as Adolf Hitler. A daily Instagram scrolling experiment found roughly 25–30% of the feed consisted of Sora videos crossposted as Reels, with watermarks that are not highly noticeable when scrolling. Most visible Sora outputs remain fantastical or obviously fake, but examples such as people or objects being sucked into tornadoes demonstrate clear potential for deliberate deception and misuse.
Read at Poynter
Unable to calculate read time
Collection
[
|
...
]