The social networking company said it will use automated technology to look at the content of a tweet and “the relationship between the Tweet author and replier” to determine if a block is warranted. Twitter previously introduced the feature during an Analyst Day presentation in February. 

If the company detects that one user is “using potentially harmful language—such as insults or hateful remarks—or sending repetitive and uninvited replies or mentions,” it will automatically block the offending user on behalf of the targeted person, Twitter said Wednesday in a blog post. Automatic blocks last for seven days. 

The goal is to stop “overwhelming and unwelcome interactions that can interrupt” a user’s experience, the company said in the post. 

Twitter has struggled for years to manage abuse on its platform, and has worked in recent years to make it easier for people to report offensive posts. Chief Executive Officer Jack Dorsey has also talked about the need to limit the burden on users who are harassed by letting other people report on their behalf, or by automating the process. 

Safety Mode is being tested for now, and Twitter said its technology can determine which users interact regularly to ensure it doesn’t block someone by mistake. Any person who uses Safety Mode can review which accounts have been automatically blocked for them and edit that list. 

More tech coverage from Fortune:

• Reddit defends allowing COVID-19 misinformation to spread on its forums
• He’s just 12 years old and has already made over $350,000 selling NFTs
• Chinese smartphone maker Xiaomi sets sights on Samsung’s throne
• The pandemic is the gift that keeps on giving for Best Buy
• Silicon Valley’s Long-Term Stock Exchange finally lists its first companies: Twilio and Asana