This is a further step in the attempt to limit the negative influence of artificial intelligence on the democratic game. On Friday, February 16, around twenty companies in the digital technology sector announced that they had reached an agreement to collaborate on “preventing misleading artificial intelligence content from interfering with elections around the world this year”, according to the Press Agency Reuters.
Two types of companies are among the signatories to this agreement, announced at the Munich Security Conference. On the one hand the developers of these generative artificial intelligence (AI) models, such as OpenAI, Microsoft or Adobe. On the other hand, social media platforms such as Meta (owner of Instagram, Facebook and WhatsApp), TikTok and X (formerly Twitter), which will have to try to prevent this harmful content from being distributed on their sites.
“Today, Meta signed the technology agreement to combat the spread of misleading AI-generated content during the 2024 elections […] We hope the industry will take a significant step to address this challenge,” Nick Clegg, president of global business at Meta Platforms, responded to X.
Avoid misleading images and sounds
Because generative artificial intelligence, which makes it possible to create images, videos, or sound documents on demand and at lower cost, is a weapon that can upset certain elections. A worrying prospect as half of the world’s population will be affected by at least one election (local, regional, national, supranational) during the year 2024, starting with the United States, the European Union, Russia or India.
Across the Atlantic, artificial intelligence is already disrupting the race for the White House. Recently, opponents of President Joe Biden spread false messages calling on voters to stay home during the Democratic primaries. As for the Republican primaries, supporters of Ron De Santis, rival of Donald Trump, shared doctored videos of the former President of the United States.
The agreement signed by these tech companies thus includes a commitment to develop “tools for detecting misleading images, videos and sounds generated by AI”, details Reuters. This could include watermarking content to identify it as AI-generated, or mentioning it in the metadata. The companies also plan to “create public awareness campaigns to educate voters” about these new types of content. However, they did not mention a timetable or explain how each would implement these objectives.
A security issue
Signed at the 60th Munich Security Conference, the agreement demonstrates the growing concern among political and business leaders that artificial intelligence is a problem for democratic elections if not well regulated. “An unregulated AI race carries risks. AI systems have already caused various incidents,” state in their 2024 report the organizers of this summit, bringing together the elite of world geopolitics.
“Realistic models of image and video generation make the problem of misinformation worse […] While policy makers [des divers pays] are focused on winning the AI race, they should not underestimate these problems,” they warn.