"AI technologies can be of great service to humanity" and "all countries can benefit", but "they also raise fundamental ethical concerns", underlines in its preamble the 28-page recommendation, ratified by the 193 Member States of Unesco.

There is "a need to ensure the transparency and intelligibility of the functioning of the algorithms and of the data from which they have been trained", because they can influence "human rights and fundamental freedoms, equality gender, democracy ", continues this international organization based in Paris.

Almost absent at the start of the millennium, AI has gradually entered our lives: it decides what news we will read on our phone, what films will be offered to us in streaming, what routes the guidance systems will take us ...

But the algorithms that make it work have also been misused in recent years, illustrating its dangers.

Facebook has been at the center of several scandals.

The British firm Cambridge Analytica has been accused of having misappropriated the data of the American giant to influence politically the referendum which led to Brexit in the United Kingdom, and the election of Donald Trump in the United States.

The result of work that started in 2018, the Unesco recommendation highlights values ​​- "Respect, protection and promotion of human rights", "diversity and inclusion", promotion of "peaceful societies" and the environment - which the member states undertake to respect.

It also lists the actions that the signatories will have to carry out, in particular the establishment of a legislative tool to frame and monitor AIs, "ensuring total security for personal and sensitive data" or even educating the masses about them.

"We wanted to have a universal and relatively precise tool, which does not remain at the level of values ​​and generalities", welcomed the Director General of Unesco, Audray Azoulay, hailing "a victory for multilateralism" at a conference Press.

"Global normative framework"

This recommendation "sets a global normative framework and gives Member States the responsibility of translating this framework at their level. It is like a source of inspiration, a shared referent", of which Unesco will then verify the progress country by country. , she added.

Russia, China and even Iran, states regularly accused of obstructing human rights, are signatories to the text which, UNESCO recognizes, is an "incentive device" without the possibility of sanctions.

Which raises questions about its scope.

Audray Azoulay in Paris on April 14, 2016, then Minister of Culture Audray MIGUEL MEDINA AFP / Archives

"If this text had no power, these countries would not even have come to discuss it," said Alessandra Sala, director of the artificial intelligence service of the content provider Shutterstock, interviewed by AFP.

They ratified the recommendation "because they realize that we are at a pivotal moment in AI and they don't want to be seen as countries heading in the wrong direction," she adds.

The United States and Israel, very active in new technologies, are not part of Unesco.

But the administration of President Joe Biden "is working on legislation" on AI, a subject on which "the United States is ahead of Europe", some American states having already greatly complicated the collection of data on their territory, remarks Mrs Sala.

Another pitfall of this text, it does not impose anything on companies active in artificial intelligence, in particular on the American internet giants, the famous Gafa (acronym for Google, Amazon, Facebook and Apple).

David Leslie, a researcher at the Alan Turing Institute on Science and AI, however apprehends it as "a step in the right direction", which other international institutions, such as the Council of Europe or the European Union will follow soon, according to him.

"Texts like this put a lot of pressure on the Facebooks of this world," whose "predatory behavior is increasingly exposed," he said.

And to launch: "To act with as much impunity as Facebook did, it is no longer possible."

© 2021 AFP