Tweets that go against the new policy will be taken down, the micro-blogging platform said in a statement on Tuesday.
"If reported, tweets that break this rule sent before today will need to be deleted, but will not directly result in any account suspensions because they were tweeted before the rule was set," said the company.
Last year, Twitter asked for feedback from different communities and cultures to ban hateful content based on religion.
In two weeks, it received more than 8,000 responses from people located in more than 30 countries.
"Across languages, people believed the proposed change could be improved by providing more details, examples of violations, and explanations for when and how context is considered," said Twitter.
Many people also raised concerns about Twitter's ability to enforce its rules fairly and consistently.
"Our primary focus is on addressing the risks of offline harm, and research shows that dehumanizing language increases that risk," said Twitter.
The respondents also asked the company to use clearer language when defining policy violations and to be more consistent in its enforcement of the rules.