The software reportedly determines a terrorism related account through how it behaves, rather than what it posts, saying the accounts have "distinctive behavior".
Most of these accounts, about 74 percent, were discovered because of the company's internal spam-fighting tools, according to a government transparency report that Twitter releases twice a year.
The company's statistics show that it had to ban over 375,000 accounts for spouting hateful rhetoric in the last half of 2016 alone, bringing the total number of accounts banned for such behavior to over 630,000. Others were suspended following outside requests, including those from government representatives.
Twitter said it received 716 government TOS requests related to promotion of terrorism from July 1 through December 31, 2016, covering 5,929 accounts, and action was taken on 85 percent of those accounts.
Over 375,000 accounts were suspended by the social media website in the last six months of 2016 in a bid to crackdown on extremism. For now, we have added the new government terms of service section to this report with information covering government requests and our response to removing content that violates the Twitter rules relating to the promotion of terrorism.
"Unless we are prevented from doing so, when we withhold content in a certain country, Twitter will continue to provide a copy of the request to Lumen so anyone can see what type of content was removed and who made the request", the company said. It also includes an update on the company's continued work to remove terrorist content from our platform beyond government reports.