Request A Demo

Protect Your Community

with Trust & Safety infrastructure you can customize.

Learn More

Our customers take online safety personally.

 

Safe, inclusive, engaged communities are not born. They are deliberately made and maintained by invested community members and passionate Trust & Safety professionals.

We exist to power their efforts.

4239
Mercarilogo
Screen Shot 2020-07-28 at 9.15.40 AM

Here's how we help.

Our Contextual AI system identifies behaviors happening in real time, across different content types and languages by evaluating many inputs, not just one word or line of text. It assembles those inputs into a larger picture and, from that vantage point, is able to make a more accurate determination of what may be happening.

Metadata

Timestamps, profile information, other conversations and content produced by involved users, and more pieces of metadata are considered.

 

 

Classifiers

Custom classifiers developed specifically for our customers are used to detect behaviors in a single piece of data - like a chat message, audio clip, or user name.

User Reputation

A customizable user reputation system scores users on attributes our customers select.

Lookups

Custom lookups bolster other behavior identification methods.

Aspect Models

Aspect models detect complex behaviors like sexual harassment, CSAM grooming, or radicalization, within larger pieces of data, like entire conversations, user streams, or forums.

Custom Embeddings

Custom embeddings are created to understand the language unique to a platform.

Explore Our Models

Improved Accuracy = Better Safety

Our customers trust the signals our Contextual AI system produces, and are able to build automated responses where necessary. Automation helps Trust & Safety professionals scale their efforts and extend safety benefits quickly and easily to all community members.
LEARN MORE

Moderate Confidently

The accurate signals our Contextual AI system produces can be fed into home-grown moderation technology or into Guardian, our moderation system.

Learn More
Guardian-GIF

The Data Vault

AI is only as good as the data it is trained from. Our Contextual AI system is trained with data from our Data Vault.

Content Types

The Vault contains many different types of data, from voice, video and text to memes, emojis and images.

Content Settings

The setting of data can change its interpretation. So, the Vault has data within a variety of settings, from posts, searches, comments, blogs, profiles, etc.

Languages

Behavior data in multiple languages, both logographic and non-logographic, is housed in the Vault to power our patent-pending approach to multi-language moderation.

Data Shapes

People communicate differently in essays, books and articles then they do in posts and messages. The Vault contains different shapes of data to account for that.

Industries

The Vault holds data from different industries including video gaming, dating, marketplaces and social media.

Learn More

Our Customers Put Their Community First

pinterest-logo

“To build a service that helps inspire people to find and do what they love, we have to deliberately engineer a safe and positive experience. That’s why we partner with Spectrum Labs.”

Rahim Daya
Head of Search
Pinterest

Take a closer look.

Screen Shot 2020-07-24 at 9.00.32 AM

Moderating Voice Chat

As Trust & Safety professionals in the gaming industry know, moderating voice chat is really hard. Accuracy, cost, speed and privacy challenges have combined to make moderating voice chat not for the faint of heart.

We’re working to solve these challenges. We’re collecting best practices and building a better way.

Learn More

Screen Shot 2020-07-24 at 9.00.37 AM

Supporting Multiple Languages

Traditionally, Trust & Safety professionals have moderated across languages by maintaining lists of translated keywords. Sure, translation software has gotten better, but the step ads delays. Don’t even get us started on the problems with keywords.

Our customers have stopped translating, embracing our patent-pending approach to multi-language support and confidently moderate across the languages used in their communities.

See How

Lessons from the front lines

On The Brief, Trust & Safety professionals share stories from the trenches, passing on what they’ve learned.

Dealing with a tough issue? You may find an episode that addresses it.

Listen Now