SAN FRANCISCO, California: OpenAI launched a new social media app on September 30 that lets users create short videos with artificial intelligence, stepping directly into the space dominated by TikTok, YouTube, and Instagram.

The iPhone-only app Sora allows people to generate clips of themselves in almost any imaginable setting, from anime-inspired dreamscapes to hyperrealistic environments. Its debut video featured an AI-rendered version of CEO Sam Altman greeting users from a psychedelic forest, the moon, and a stadium of rubber duck races before passing the spotlight to colleagues in other surreal scenes. For now, Sora is available in the U.S. and Canada.

Meta rolled out a similar feature last week, launching its Vibes short-form AI video feed within the Meta AI app. CEO Mark Zuckerberg announced it with a series of whimsical AI-generated clips, from a cartoon version of himself to a kitten kneading dough. Both products promise personalized feeds tailored to users' engagement.

The sudden push from two of the biggest AI firms signals a new front in the battle for attention online. But critics worry the result will be an endless wave of what some call "AI slop" — content that looks convincing but risks overwhelming genuine human creativity and trust in information.

"These things are so compelling," said Jose Marichal, a political science professor at California Lutheran University who studies AI's social impact. "What sucks you in is that they're kind of implausible, but they're realistic looking."

Marichal said his feeds are already packed with such videos, from cats riding wild animals to fake natural disaster reports that spread quickly despite being easy to debunk. "We need an information environment that is mostly true or that we can trust because we need to use it to make rational decisions," he said. Without that, he warned, people risk becoming "either the manipulated or the manipulators," undermining democracy itself.

OpenAI acknowledged those concerns in a blog post. It said the app is designed with guardrails to reduce doomscrolling and isolation, pledging to survey users about their well-being and bias feeds toward posts from friends rather than strangers.