WASHINGTON, United States (AFP) — Facebook, Twitter, and YouTube were pressed in Congress Wednesday over their reliance on artificial intelligence and algorithms to keep their powerful platforms clear of violent extremist posts.

In a Senate Commerce Committee hearing, executives of the world’s top social media companies were praised for their efforts so far to eliminate Islamic State, Al-Qaeda and other jihadist content from the internet.

But critics say that extremist groups continue to get their propaganda out to followers via those platforms, and call for tougher action.

Another concern is that the continued ability to use anonymous accounts, while benefiting pro-democracy activists battling repressive governments, will also continue to empower extremists.

“These platforms have created a new and stunningly effective way for nefarious actors to attack and to harm,” said Senator Ben Nelson.

The current efforts by the companies to remove content and cooperate with each other in doing so are strong but “not enough,” he said.

YouTube is automatically removing 98 per cent of videos promoting violent extremism using algorithms, said Public Policy Director Juniper Downs.

But Senator John Thune, Chairman of the Commerce Committee, asked Downs why a video which showed the man who bombed the Manchester Arena in June 2017 how to build his bomb has repeatedly been uploaded to its website every time YouTube deletes it, as recently as this month.

“We are catching re-uploads of this video quickly and removing it as soon as those uploads are detected,” said Downs.

Carlos Monje, director of Public Policy and Philanthropy for Twitter, said that even with all their efforts to fight terror- and hate-related content, “It is a cat-and-mouse game and we are constantly evolving to face the challenge.”

“Social media companies continue to get beat in part because they rely too heavily on technologists and technical detection to catch bad actors,” said Clint Watts, an expert at the Foreign Policy Research Institute in the use of the internet by terror groups.

“Artificial intelligence and machine learning will greatly assist in cleaning up nefarious activity, but will for the near future fail to detect that which hasn’t been seen before.”

Last year Google, Facebook, Twitter and Microsoft banded together to share information on groups and posts related to violent extremism, to help keep it off their sites.