Roblox rolls out AI tool to combat predators amidst numerous child abuse lawsuits


Roblox is releasing a new open-source tool powered by artificial intelligence to help detect predatory language in game chats. This comes after numerous lawsuits accusing Roblox of turning a blind eye to child exploitation on its platform.

The virtual universe, popular amongst young people, has received a negative reputation lately due to the numerous lawsuits alleging that it has chosen to ignore predatory behaviour on its site.

To combat this and perhaps reverse its negative image, Roblox is rolling out an open-source AI tool, Sentinel, that helps to identify the warning signs of possible child exploitation.

ADVERTISEMENT

The tool attempts to detect sexually exploitative communications “long before something becomes explicit,” as Sentinel allows Roblox “to detect and investigate subtle patterns early,” the company said in a blog post.

Cybernews has reached out to Roblox for comment.

Sentinel has been running on the platform since 2024 and proactively detected 35% of cases in the first half of 2025.

“In the first half of 2025, its proactive capabilities have helped our team submit approximately 1,200 reports to the National Center for Missing and Exploited Children,” Roblox continued.

Niamh Ancell BW jurgita Paulina Okunyte Konstancija Gasaityte profile
Don't miss our latest stories on Google News

How does Roblox’s AI tool work?

Sentinel works by capturing one-minute snapshots across six billion in-platform messages.

These chats are then analyzed using machine learning to identify harmful language and violent intent, such as grooming or child endangerment.

ADVERTISEMENT

“Additionally, we aggregate this information over time, identifying concerning cases and patterns for human analysts to assess and investigate,” Roblox said.

Roblox claims that, unlike other tools that “rely on static rules and labeled examples,” Sentinel self-teaches based on communication observed in the platform.

This allows the AI tool to make assessments based on normal communication and thus spot abnormal communication as it arises.

The Roblox team did this by “developing two indexes,” one based on normal, safe conversations between users and the other comprised of flagged content removed due to violations of its child endangerment policy.

Sentinel works on just 13,000 examples of negative behaviour, which Roblox claims is enough to successfully identify potential harm.

Roblox hit with countless child-endangerment lawsuits

The “child-friendly” platform has recently been hit with countless lawsuits alleging child endangerment.

Cybernews previously reported on a case involving an 11-year-old Jane Doe who was coerced into engaging in sexual acts with a pedophile she met on Roblox.

Another example of child endangerment presented in the same lawsuit depicted a young girl who was virtually raped by two male avatars in the virtual space, specifically in a “playground.”

In a more recent lawsuit reported on by Cybernews, a 10-year-old girl was coerced into giving a Roblox predator her physical address and was subsequently kidnapped by the abuser she met on the platform.

ADVERTISEMENT

The 27-year-old drove to the young girl’s home and kidnapped her, holding her “against her will for an extended period.”

The lawsuit alleges that the man and the child stayed in a hotel parking lot after being denied a room.

FAQ: What’s going on with Roblox?

Update: Matt Kaufman, Chief Safety Officer at Roblox, told Cybernews that “AI has become foundational to how Roblox approaches safety, and how we help keep our users safe. Contributing to the open-source community is key to our approach to online safety as we know these challenges exist across the digital world. We aim to share our technologies and collaborate with other leaders to make the internet a safer place.”