ActiveOS Updates: Underage Detection, Policy Management, and More

Over the past month, our product team has been busy enhancing ActiveOS and ActiveScore with new AI models, greater customization, and more. Our latest updates help protect young users, manage policies efficiently, and detect emoji abuse. Here’s a quick overview:

Protect your Most Vulnerable Users with Accurate Underage Detection

As more of children’s daily lives are spent online, their safety is threatened in multiple ways, and regulators are taking action. 

Legislation like the US Children’s Online Privacy Protection Rule (COPPA), the EU’s Digital Services Act (DSA), and the UK Online Safety Bill requires platforms to protect minors from online harm and ensure their privacy. While each law has its own focus, a core component of all of them is accurately assessing whether or not a user is a minor. 

ActiveScore’s improved underage detector now accurately assesses users’ age in text, images, and videos – regardless of the self-reported age. With a single integration, chats, images, and other media types are analyzed, producing a binary result whether the user is under 18. Using similar queues, the model can also assess whether the user is under 13, or between 13 and 18 years old. 

Automated Actions for Legal Compliance
Use ActiveOS’ policy management tool to combine the underage detector with models for CSAM, nudity, alcohol, illegal goods, and more. Drag-and-drop indicators to create automations that enforce these policies, such as limiting access by age, or reporting illegal content to authorities.

Continuous Improvement
ActiveScore’s underage detector is built into our Child Safety models, providing extensive coverage against policy violations. Our models continuously improve through moderator feedback and retraining to align with real-world changes, new intelligence findings, and unique policies.

Benchmark Information:

Based on an analysis of 24K+ items from a live-streaming platform.

Customized Detection with Categories

ActiveOS Policy Management tool allows combining detectors like ActiveScore models, custom AI models, and keywords for tailored policies.

While previously, every piece of content was analyzed against all activated policies, our new categories feature enables customized detection by setting specific parameter triggers for each policy – resulting in greater accuracy and significant cost savings due to fewer API calls being made. 

How it works
Categories can created based on anything passed in an API call. For example, using the underage detection API above may result in a category of “under 13” users, for which teams can set specified policies. Other parameters for categories can include media types – for example setting unique policies for usernames, chats, and comments.

When creating a new policy, users can now choose to enforce it on:

• One Category: Policy runs only on the specified content category.
• Multiple Categories: Policy runs on each specified content category.
• No Category: All policies without categories will be run. If an item has a category, it won’t trigger that policy.

Here you can see a few examples:

Detect Emoji Abuse in Keywords

We know that threat actors frequenly use emojis and combinations of emojis to evade detection. That’s why we’ve updated our keyword tool to now support emojis. Add emojis to keyword lists just like you would any other character or phrase. Use fuzzy match to combine emojis with other emojis, or in combination with other letters. 

For broader protection, keyword lists can be combined within the policy management tool to build more robust policies. Learn more about our policy management tool here. 

Stay tuned, as we are always working on more exciting features and enhancements for ActiveOS and ActiveScore. If you’d like to see these new features in action, learn more about these and other features, feel free to schedule a 1-1 demo session with us. 

Thanks,

The ActiveFence Team