Kim Kardashian deepfake (Apr 2025)

In April 2025, Roblox chief executive David Baszucki publicly backed the Take It Down Act, US legislation requiring platforms to remove non-consensual deepfake content. Coverage of the endorsement pointed straight back to one of the platform's most notorious moderation failures: a user-created Roblox experience that advertised a fake explicit video of Kim Kardashian, and that surfaced on the tablet of her then six-year-old son.

What happened

The underlying incident became public in April 2022, when an episode of Hulu's The Kardashians showed Kardashian's son Saint West bringing her a tablet displaying an in-game ad. The ad, running inside a user-created Roblox experience, claimed to promote a new explicit tape of Kardashian. No such video existed, and Roblox said the referenced footage was never actually available on the platform, but the ad itself had made it onto a service used heavily by children. Kardashian's lawyers threatened legal action, and Roblox removed the experience, banned the developer behind it, and apologized, citing a zero-tolerance policy for sexual content, as reported by Kotaku.

Three years later, Gameranx reported that Baszucki invoked the dangers of unmoderated sharing in voicing Roblox's support for the Take It Down Act, which would require platforms to remove non-consensual deepfake content. "The ability to rapidly share images and videos is one of the reasons the internet can be a complex and dangerous place for children," he said.

Why this incident matters

Roblox's audience skews young, which raises the stakes of every moderation gap. This incident shows how a celebrity's likeness can be exploited as bait even when the promised content does not exist; the fabricated claim alone was enough to spread, cause harm, and reach a child of the person targeted. It also illustrates the scale problem: platforms built on user-generated experiences and user-purchased ads cannot manually inspect everything that goes live. That is why the policy conversation has shifted from voluntary moderation to statutory duties, with the Take It Down Act obliging platforms to remove non-consensual intimate imagery, including AI-generated deepfakes, once it is reported.

Where detection fits

Manual review does not scale to millions of daily uploads, which is where automated screening earns its place. Running new video assets and ad creative through a video deepfake detector lets platforms flag probable face swap fabrications of real people before they are served to users, and gives trust-and-safety teams evidence to act on when a takedown request arrives.