San Francisco (AFP)

Twitter announced Tuesday that it would start fighting "falsified" photos and videos in the wake of other social networks, summoned to take responsibility, especially during the campaign for the US presidential election.

The platform intends to focus on modified content (video or audio montages, edited images) which aim to deceive the public or risk harming people, by inciting violence or by infringing their freedom of expression, for example .

Tweets falling into these categories will be removed or tagged with a warning, starting in March. The network can also reduce the visibility of messages or add context.

Most major social networks have implemented measures combining artificial intelligence and human resources to fight against disinformation, from fake news to "deep fakes" (fake photos or videos, hyper-realistic, editor's note).

They react in particular to pressure from European and American authorities, while manipulation campaigns carried out notably on Facebook in 2016 tried to influence opinion during major elections, such as the presidential election in the United States or the referendum on Brexit in the UK.

"This new rule is in addition to the many other existing rules" to regulate Twitter, said Yoel Roth, responsible for the integrity of the platform, during a press conference. "For example, for years we have been preventing the spread of images and videos of fabricated or falsified sexual content, which is very widespread on the internet".

The tweet network targets rigging, including audio or video, but does not directly attack spurious written messages, while it has also banned advertisements of a political nature.

A video montage, like the one which lent to the democratic candidate Joe Biden of the racist remarks at the beginning of the year, should no longer be cited on Twitter. One of the sharing tweets in this video had been viewed over a million times.

- Two weights, two measures -

"Whether you use advanced machine learning tools or an inexpensive app to slow videos down, our rules will apply to content, not how to make it," said Yoel Roth.

To identify potentially problematic content, Twitter teams around the world rely on reports, even if "we want to reduce the burden on users," said Yoel Roth.

He admitted that the satirical images and videos could give them a hard time. "If we are wrong, there will be an appeal procedure," the network said.

Google’s video platform YouTube announced similar measures on “manipulated or falsified content” on Monday “in order to deceive users” and “who pose a flagrant risk of harm”.

Facebook, for its part, still allows political ads, and even exempts them from its fact-checking system.

AFP participates in more than 30 countries and 10 languages ​​in the "Third party fact-checking", a media verification program developed by Facebook. With this program, which started in December 2016, Facebook pays around sixty media around the world, general and specialized, for the use of their "fact-checks" on its platform and on Instagram.

On the dominant social network, if information is diagnosed as false or misleading by one of these partners, users are less likely to see it appear in their news feed. And if they see it or try to share it, the platform suggests that they read the verification article. There is no deletion of publications. The participating media are completely free in the choice and treatment of their subjects.

© 2020 AFP