The Solution

Lopez turned to ActiveFence to enhance the safe browsing experience. “While our main goal is to provide relevant results to user queries, we also need to offer a safe searching and browsing experience,” says Lopez. “That’s where ActiveFence came into play.”

After a short integration process, ActiveFence implemented its intel-fueled, AI-powered detection models to scan search results and identify any violative CSAM or adult content across its URLs, images, or queries. This included multilingual coverage, including detection in English, Spanish, German, and French.

One of the models is ActiveFence’s novel CSAM model. Trained on over six years of in-house domain expertise, this model went beyond simple hash-matching to detect novel CSAM violations, as well as multilingual terms and specific keywords or emojis used to evade detection.

CSAM and other violative content is then sent to ActiveOS in real time, where it is filtered into customized moderator queues by priority, risk and language.

“The filters in ActiveOS provide all the criteria I need to give me a better control of the data I moderate,” says Lopez.

This includes creating custom fields to determine if the content is a URL, domain, or image. He also uses custom queues to tackle the highest risk violations swiftly, such as flagged content, or CSAM-related violations.

"Moderation has become easier since working with ActiveFence,” says Lopez, highlighting how straightforward the content review process is on the platform. This has helped him scale moderation efforts, without increasing his team size.

The Impact

Since implementing ActiveFence, Brave Software has seen a reduction in high-risk harms across multiple languages. Lopez has also managed to scale moderation efforts, providing wider coverage, without spending more resources or time.

"One of the main advantages is actually saving time,” says Lopez. “Having an easy-to-use platform where I can easily check results, with fewer false positives to weed through, helps make my work more efficient.”

By continuously working with ActiveFence, Brave Software has been able to detect multilingual violative content faster, something they wouldn’t have been able to with their limited in-house linguistic capabilities.

“Over 95% of items are accurate,” says Lopez, who says the confidence of detection has helped improve moderation efficiency, and provide his users with a safe browsing experience.