Twenty digital giants pledged Friday to fight content created by artificial intelligence aimed at misleading voters, with major elections planned across the planet in 2024. These companies, including Meta, Microsoft, Google, OpenAI, TikTok and X, promise to
“deploy technologies to counter harmful content generated by AI”
, in a text unveiled on the sidelines of the major Munich security conference (MSC).
They undertake in particular to
“work on tools”
allowing them to
“spot”
misleading content created by artificial intelligence in order to identify them as such for users and control them.
To discover
PODCAST - Listen to the latest episode of our Tech Questions series
One idea to achieve this would be to put a
“digital watermark”
in videos generated by AI tools developed by these companies, invisible to the naked eye, but able to be detected by a machine.
“All solutions have limits
,” warn the signatories, among whom we also find Adobe, LinkedIn, Amazon and IBM.
“We have a responsibility so that these tools do not become weapons in elections
,” explained Brad Smith, vice chairman of the board of directors of Microsoft, which invested in OpenAI, the creator of ChatGPT.
So-called generative AI makes it possible to create, upon simple request in everyday language, text, images, sound files or videos, capable of passing off documents created from scratch as authentic.
Also read: Sundar Pichai, boss of Google: “Paris is a magnet for tech talent”
Beware of tricks
Thanks to archives, the technology can in particular produce
“deepfakes”
, documents which show a person saying or doing something when it never happened.
Digital giants are in a hurry to act in the face of the spread of this content, while major elections are planned for 2024 around the world, notably in the European Union, the United States, the United Kingdom, and India. and in Russia.
“With so many important elections happening this year, it’s critical that we do what we can to prevent people from being misled by AI-generated content
,” said Nick Clegg, head of Meta.
Several of these
“deepfakes”
– real tricks made using artificial intelligence – have made headlines in recent weeks, including a fake telephone message from US President Joe Biden before the New Hampshire Democratic primary at the end of January.
In Pakistan, former Prime Minister Imran Khan's party also used AI to generate speeches from their jailed leader.