Our AI based Moderation Engines are powered by real-time AI classifiers, which are designed to accurately block harmful messages before they can reach their intended recipient. All of our models can be used in parallel without sacrificing latency or accuracy, allowing for up-to-date protection without slowing down the system. Furthermore, if a harmful message is detected, you are granted full control over the action that will be taken. This includes (but is not limited to) the ability to delete or flag the message for further review (see 🔀Moderation actions). With our solutions, you can have peace of mind knowing that your system is safer and more secure.

Shields your platform against commercial spam, like users or bots promoting goods, services, or scams in the chat.

Commercial Spam (cSpam) is User Generated Content (UCG) which aims to promote, advertise, solicit or direct recipients to take a specific action, usually for commercial reasons, including fraud, job requests/offers, financial transfers, begging, advertising, or scams. cSpam may ask users to visit a specific URL, it may repeatedly mention a product, company, person, trademark, or other identifier without regard for the context in which it is posted, or it may be in the form of similar but not identical content.

CSpam is a major problem for users and event organizers alike, as it requires a specific and targeted effort to be managed properly. In live events, cSpammers send unsolicited invites and promotional content without permission, which can be especially disruptive and overwhelming to other users. This type of cSpam can take the form of links to profiles in both live events and livestreaming, and can be anything from self-promotional messages to aggressive attempts to direct users to external websites. Not only does this create a chaotic and disruptive environment, but it also can be a significant drain on the resources of event organizers, who must take extra steps to prevent and manage cSpam in order to create a safe and enjoyable experience for all.

the cSpam engine can be used to detect such messages in real time, and encourage users to limit the amount of solicitation content they send to others to reasonable bounds.

In B2B and B2C marketplaces, unauthorised advertisers contacting sellers through the sellers’ listings is an issue which may lead to sellers disengaging from the marketplace. These unauthorised advertisers are not interested in purchasing the items listed, but are instead using them as a way to contact the seller and spam them with advertising material.

Stream’s cSpam engine can differentiate between messages looking to advertise or sell, and messages asking for product information, or purchase requests. By using the Roles and Permissions functionality, you can chose to apply cSpam detection only to those contacting sellers through markeplace listings, to reduce the amount of commercial spam messages these sellers encounter.

Unprotected dating platforms will eventually encounter phisers. users will be targetted by scammers who will engage users in conversations, and eventually try to sell them something or direct them towards a scam.

The cSpam is effective at detecting users sending others solicitations of a commercial nature, and can be used to warn users engaging in this behaviour, and educate them about platform rules. If they persist in breaking the rules, then those users can easily be banned.

In crypto communities, ‘shilling’ describes the phenomenon of users, frequently new to the platform, enthusiastically evangelising novel cryptocurrencies to users, without engaging with the existing community.

cSpam can effectively be used to manage and mitigate the impact of ‘shilling’ by blocking new users from sending these sorts of messages to other users.

In MMO games and similar gaming communities, some users may spam chat channels with messages looking to sell in-game items for real world currency.

cSpam protection can detect these messages, and can be used to flag ‘gold-farmers’ for review by moderation teams.

Go to Settings > Channel Types

Click on the desired channel type

Click on “Commercial Spam” > Configure > select the desired action

• Hi there, my name is John with the Ipsagon team. Are you interested in learning more about our top rated solution in Application Performance Monitoring?

• Need money for college? Live in Springfield County? Have a 2.5 GPA or higher? Potsdam Valley Alumnae Delta Sigma Theta may have a scholarship for You! Visit us in the Expo link.

• Only for my followers: on Black Friday (27 Nov) I can grant you a 10% discount on this reference! Let me know if you are interested. Thanks

• Interested in learning more about the brain & its marvelous ways? Then the Neuroscience Research Group (NRG) is the right club for you! Don't forget to join our booth to learn more about our club!

• Click this link for a chance to win 10000$! www.scam-site.com

Shields your platform against attempts at taking transactions and conversations outside of it.

Platform Circumvention is User Generated Content (UGC) which attempts to direct recipients off-platform in situations where that is not allowed by the platform. This is common in marketplaces that process transactions, and other scenarios where a platform plays a safeguarding role such as dating, gaming and social.

In some cases, they may request a recipient's email address or phone number to bypass a three-strike system. Alternatively, they may communicate their Instagram name, Venmo/CashApp details, or direct users to search for their profile on a different platform or go directly to their website to make a purchase.

For many marketplaces or on-demand providers, revenue and customer safety relies on the transaction happening through the platform. Thus, scammers and other bad actors may attempt to direct users off-platform to conduct transactions out of the safety of the platform itself. This can cause harm to users, and the platform’s bottom line.

Marketplace platforms often attempt to block messages containing phone numbers, email addresses, or other contact information using regex checks. To avoid detection, bad actors may use unicode characters or substitution characters to convey their contact details and lure recipients off-platform. This makes it hard to process messages using text-based filters, and can lead to long processing times, resulting in a degraded user experience for all users.

The Platform Circumvention engine can detect attempts at taking transactions off-platform, and can be used to block messages requesting contact details, or directing users to other platforms. This works even when users try to mask this behaviour.

Dating platforms are increasingly being used as ways to advertise and direct users off-platform, onto other platforms which host commercial content. This can harm user engagement levels and result in declining engagement and retention rates, harming the long-term viability of dating platforms.

Platform Circumvention detection can detect the messages used to direct users off-site, even when they are obfuscated using symbols and non-standard characters.

In games and gaming platforms with users under the age of 18, platforms may want to prevent children from sending their name, address and other contact details to others on the platform.

Platform Circumvention protection can prevent this scenario from happening, and can be used to prevent users with specific user roles from sending out their contact information, or can be used more broadly to detect users soliciting others for this information as well.

Social networks built on anonymous or pseudonymous usage or which may have a legal obligation to prevent the transmission of personally identifiable information of certain user groups on the network can use Platform Circumvention protection to detect when this is happening, and inform users of the rules when needed.

Similarly to cSpam:

Go to Settings > Channel Types

Click on the desired channel type

Click on “Platform Circumvention” > Configure > select the desired action

• hi, here is my pay.pl account 1239 543- three-844

• Hi ♥️fullset Thank you for interesting on item ♥️ For more items in🧡st🎀gr🧸m Coco5000 💕

• hey, quiero vender el producto por menos cantidad pero esta app se lleva mucha comisión y por eso elevo la cantidad. quieres que hablemos por e,m,a,i,l y asi yo obtengo lo que quiero y tu pagas menos? te puedo enviar serial, factura, etc

• Hi ✨ I’m @kira9309273. If you are interested in this item, I’m flexible and open to an offer. Please feel free to contact me😊 Thank you💝 Kira🍀

• Hi yes are you still looking for a job if yes Dm on this number three six zero four four seven eight four seven one Or send your number in w o r ds

• Bonjour vraiment très jolie l’article !!j emerais avoir d'autres photos pas Whats// ssap! Pour que je le montre as ma fille merci je te laisse mon numéro : Zéro 6 Quarante-trois 69 Cinquante-quatre 47

• Coucou bonjour votre article est toujours disponible et je suis intéressé par votre article et je peux avoir une photo visible sur mon numéro ... zéro 6…....... cinquante cinq55........ quatre vingt cinq85…...... onze11.......... quarante six46

• Hi! Would you be interested in finding a deal on another platf.orm?

Detects a message's intended meaning, rather than relying only on exact word matches.

AutoMod semantic filtering is an AI-based moderation tool that can detect and action incoming messages in real time, based on their meaning and intent. It does not require any software development, data science or machine learning skills to use. Instead, you provide the tool with a list of example phrases corresponding to the meaning and intent you want to detect, and it will detect messages with similar meaning and intent to those example phrases, even across languages. You can change the phrases at any point through the moderation dashboard, and the changes will be reflected almost immediately.

This engine allows Trust & Safety specialists, Content Moderators, Community Managers, and other moderation professionals to implement their specific, fine-grained content moderation policies in an automated manner, without having to resort to either costly and time consuming training of machine learning models, cumbersome Regular Expressions or broad word filters.

The tool is designed to moderate based on the intent, rather than pattern matching or exact word patches, and makes it simple for moderators to autonomously update and action changes in their organisation’s moderation policies and react to emerging challenges as soon as they appear.

Go to Settings > Channel Types

Click on the desired channel type

Click on “Semantic Filters” > Configure

You will be able to select phrase lists

Check out our user tutorial to get started.

Configured phrase:

Darth Vader is Luke's father

Messages recognized:

• Darth Vader reveals to Luke that he's his father

• Luke Skywalker is Darth Vader's son

• Luke Skywalker was fathered by Darth Vader

• "I'm your father!" said Darth Vader to Luke Skywalker

As part of Stream basic moderation, we offer possibility to create blocklists that allow to filter messages based on certain words. For more info, see blocklist documentation

Go to Settings > Channel Types

Click on the desired channel type

Click on “Blocklists” > Configure

You will be able to select a blocklist to enable, and configure it to either flag or block within the scope of this channel type.