Twitch’s new tool uses machine learning to spot users evading bans

2 Dec 2021

Image: © ink drop/Stock.adobe.com

The new AI tool flags suspicious accounts to help streamers keep abuse and harassment from their channels.

Twitch has launched a new tool powered by machine learning to help its users detect people who are trying to bypass channel-level bans.

The game-streaming platform said this new tool is part of an ongoing mission to keep the chat sections of user channels free from hate and harassment.

While users can already ban accounts permanently from their chat, Twitch said that people often try to create new accounts so they can go back into the chat and continue abusive behaviour.

The new Suspicious User Detection tool analyses a number of account signals using machine learning to spot users who are trying to dodge bans by making new accounts.

This tool will flag suspicious accounts as either ‘likely’ or ‘possible’ channel-ban evaders and notify the owner and moderators of the channel.

If an account is flagged as ‘likely’, messages won’t be sent to the chat and will only be visible to the creator and moderators, who can choose to leave the restriction as it is, monitor the user or ban them from the channel.

If it’s a ‘possible’ ban-evader, the message will go into chat normally but the account will be flagged for creators and moderators so they can monitor the user and restrict them from chatting if necessary.

The tool is activated by default for channels, and streamers can choose to disable it if they wish. Users can also manually add accounts that they want monitored more closely.

Twitch noted that machine learning is never 100pc accurate and could create false positives or negatives, which is why Suspicious User Detection doesn’t outright ban accounts that it flags. It is designed to learn from the actions of users and should improve over time.

“This tool highlights our overall approach to safety technology: build powerful tools that work together to give you finer control over your community. Our work will never be finished and we’re continuing to develop more tools to prevent hate, harassment and ban evasion on Twitch,” the company wrote.

Twitch faced a security breach in October, when a massive hack publicly leaked more than 100GB of its data on the internet.

Don’t miss out on the knowledge you need to succeed. Sign up for the Daily Brief, Silicon Republic’s digest of need-to-know sci-tech news.

Leigh Mc Gowran is a journalist with Silicon Republic

editorial@siliconrepublic.com