Less than two minutes. This is the time it took for the Twitch platform to interrupt the live broadcast of the Buffalo shooting on Saturday, which did not prevent extracts from circulating.

• Read also: The Bills and the NFL donate $400,000

• Read also: In the United States, the firearms market has exploded in twenty years

Despite advances in technology, preventing the broadcast of live violent images is still a challenge, especially since the legal framework is almost non-existent.

“If (the platforms) offer live, they expose themselves to retransmit a certain number of rapes, murders, suicides and other crimes”, argues Mary Anne Franks, professor of law at the University of Miami. "It's part of the package. »

The hundred seconds it took for Twitch to identify and take offline the "livestream" of Payton Gendron, who killed ten people in a supermarket in upstate New York on Saturday, testifies to an increased responsiveness.

In March 2019, it took Facebook 17 minutes to stop Brenton Tarrant's attack on two mosques in Christchurch, New Zealand, which killed 51 people.

Worse, in October 2019, also on Twitch this time, the author of the attack on a synagogue in Halle (Germany) was able to simultaneously upload his journey for 35 minutes, before being disconnected.

If the major social networks claim to track these videos with the help of artificial intelligence, but also of dedicated teams, the images can be quickly downloaded, edited, and posted on other sites ready to host them.

On Wednesday, a site offered several excerpts from the shooting, including a 90-second long one, viewed nearly 1,800 times since Sunday, which was removed later in the day.

The circulation of violent content beyond live broadcasts is almost systematic, for lack of applicable texts to prevent it.

“In the United States, posting a live video (from Buffalo) is not illegal,” says Ari Cohn of the TechFreedom think tank. "It does not fall into a form of expression that is not protected" by the US Constitution.

perpetual hunt

As for more established sites, spotting this violent content, often put back online in a new format, a new title, is a perpetual hunt, explains a Facebook spokesperson.

Twitter has a policy of suspending the accounts of suspected attackers and also allows itself to "remove tweets that disseminate manifestos or content" produced by these authors.

In a conversation with reporters on Tuesday, Meta's vice president of content integrity, Guy Rosen, said filters need to be calibrated carefully so as not to risk redacting related images, such as news videos or testimonials from people condemning the attacks.

Despite the investment of major platforms, preventing someone from live broadcasting even a few seconds of a violent act remains, by definition, impossible.

“The main problem is when tech groups decide, for the general public”, that live “is a tool whose usefulness exceeds the disadvantages”, advances Mary Anne Franks.

However, at a time when video is the main engine of growth for social networks, the "live" does not seem threatened, especially with the emergence of direct sales, a new mode of online commerce which has the wind in its sails.

Twitch, Facebook, Twitter, Instagram or YouTube all offer live, only Snapchat being an exception.

In writings attributed to him, the alleged Buffalo shooter had described the possibility of being able to broadcast his attack live as a “motivator”.

In the absence of federal legislation, it is up to the American states to take the initiative on the subject.

In Texas, a law ratified last September aims to prohibit social networks from dismissing content based on the “point of view” of their author. It has been criticized as likely to limit moderation and allow the circulation of violent messages or images.

“The recent tragedy (in Buffalo) underscores that this is not a partisan issue,” insists Matt Schruers, president of the computer and communications industry association (CCIA).

“Hindering the industry when it wants to go after bad actors has life and death consequences. »