Google, Meta, OpenAI and should take place this year.
The group of companies signed the Tech deal to combat misleading use of AI in 2024 elections at the Munich Security Conference held last weekend. It outlines a “voluntary framework of principles and actions” covering the prevention, detection, response, assessment and identification of the source of misleading electoral content generated by AI.
It also plans efforts to educate the public on how to protect themselves against manipulation by this type of content. The joint statement published by the signatories of the technological agreement includes the label of TikTok, Amazon, IBM, Anthropic and Microsoft.
8 commitments
With this agreement, the 20 organizations promise to respect eight points, including:
- Seek to detect and prevent the spread of misleading AI-related election content
- Be transparent with the public about how they process this content
- They will also work together to develop and implement tools to identify and slow down the dissemination of this content, as well as to trace its origin.
These efforts may include developing classifiers or provenance methods and standards, such as watermarking or signed metadata, and adding machine-readable information to AI-generated content. The eight commitments will apply to the services provided by each company.
Search models are not covered by the agreement
The deal covers content defined as “compelling” AI-generated sounds, videos and images that “deceptively simulate or alter the appearance, voice or actions” of political candidates, operatives election officials and other key players in an election, or who disseminate fraudulent information to the public about where, when and how to vote.
“In 2024, there will be more elections than ever before, with more than 40 countries and more than four billion people choosing their leaders and representatives,” the tech deal says. “At the same time, the rapid development of AI creates new opportunities as well as challenges for the democratic process. All of society will need to take advantage of the possibilities offered by AI and take new measures together to protect elections and the electoral process during this exceptional year.”
The agreement aims to set expectations for how signatories will manage risks arising from misleading AI election content created through their public platforms or open founding models, or distributed on their social and publishing platforms . These expectations are consistent with the policies and practices of the signatories.
Models or demonstrations intended for research purposes or primarily for professional use are not covered by the agreement.
“Not just a technical challenge”
The signatories added that AI can be used to counter malicious actors and enable faster detection of deceptive campaigns. AI tools can also significantly reduce the overall cost of defense, allowing small businesses to implement robust protections.
“We are committed to doing our part as technology companies, while recognizing that the deceptive use of AI is not only a technical challenge, but also a political, social and ethical issue, and we hope that others will similarly commit to action across society,” the signatories said. “We affirm that protecting election integrity and public trust is a shared responsibility and common good that transcends partisan interests and national boundaries.”
Christoph Heusgen, president of the Munich Security Conference, said the agreement was a “crucial step” in promoting electoral integrity and societal resilience. It will also help create “trustworthy technology practices.”
According to the Global Risks Report 2024 released last month by the World Economic Forum (WEF), risks associated with AI-powered misinformation on societal cohesion will dominate the landscape this year. The report cites disinformation as the top global risk over the next two years, warning that its widespread use and the tools to disseminate it could undermine the legitimacy of new governments.
Source: “ZDNet.com”