Roblox Shares AI Safety Tools to Protect Kids Online

A major gaming platform just made its most powerful child safety technology available to the world, and it could change how kids stay protected across the internet.

Roblox announced it's sharing three updated AI safety models with other companies through an open-source community called ROOST. The tools are designed to catch the warning signs that matter most: attempts to get personal information, requests to move conversations to other apps, and inappropriate voice chat involving children.

The timing matters. Nearly three-quarters of Roblox's 123 million daily users are under 18, making it one of the largest virtual playgrounds for kids. Earlier this year, a Nebraska man allegedly used the platform to contact two girls, ages 12 and 14, before moving conversations to Snapchat in what authorities described as online grooming.

The new AI doesn't just search for obvious red flags anymore. Version 2.0 looks at entire conversations instead of individual messages, catching predators who misspell app names, split contact information across multiple chats, or use coded language. The system now works in 189 languages, up from just 17.

One tool specifically watches for attempts to move kids off the platform. That shift to another app often marks the moment when safety protections disappear and danger increases. The AI recognizes phrases like "add me on" or requests for usernames on other social media, even when spelled creatively to avoid detection.

Roblox already uses these systems on its own platform. Now companies like Google, OpenAI, and Discord can adapt them for their services through ROOST, which focuses on helping smaller organizations access safety technology they couldn't build alone.

The Ripple Effect

This matters beyond just one gaming platform. Kids jump between apps constantly during a single conversation. A chat might start in a game, move to a messaging app, then land on social media. Each platform has different protections, creating gaps predators can exploit.

When safety technology gets shared across platforms, those gaps start closing. A predator's tactics might get flagged on multiple services instead of just one. Smaller apps and games that serve children but lack massive tech budgets can now access tools designed to catch sophisticated threats.

The company reports its accuracy score jumped from 63.41 to 90.52 on standard safety measurements. That means fewer dangerous messages slip through while fewer innocent conversations get wrongly flagged.

Parents still need to stay involved. This technology helps, but no AI replaces having direct conversations with kids about online safety, monitoring who they talk to, and knowing which apps they use.

More companies protecting children with shared intelligence means safer digital spaces for the next generation.