Officials from the Group of Seven leading democracies have broadly agreed on a draft of guiding principles for developers of advanced forms of artificial intelligence, hoping to mitigate associated risks such as harassment, disinformation and privacy concerns.
The G7, involving Japan, the United States and the European Union, has been discussing the need for international standards to harness rapidly developing technologies like AI bot ChatGPT.
During a meeting in Kyoto on Oct. 9, G7 representatives, including Japanese Minister of Internal Affairs and Communications Junji Suzuki, agreed on 11 principles. These include taking appropriate measures throughout the AI lifecycle, such as independent external testing to address risks and vulnerabilities, as well as steps to protect copyright.
Their leaders are expected to endorse the principles before relevant ministers reach a formal agreement later this year.
The meeting was held on the sidelines of the Internet Governance Forum, an annual U.N. conference.
The G7 has warned of potential risks posed by AI, including online harassment, hate and abuse, and threats to children's safety and privacy, as well as the danger of information manipulation.
Japan, which chairs the G7 this year, has been leading the group's talks on rules to make advanced AI systems safe, secure and trustworthy as AI is considered essential in bringing innovation for solving global challenges like the climate crisis.
One of the draft principles concerns the development of watermark tools, designed to mark and identify contents created by AI by concealing electronic identifiers in images and other content as part of efforts to counter misinformation and misuse of AI-generated content, according to the European Commission.
In line with the G7 principles, Japan is developing domestic guidelines to reduce the risk of overreliance on AI systems and technologies.
The G7 also aims to compile an international code of conduct for AI tool developers to prevent relevant technologies from being used in crimes, as well as guidelines for AI service providers and users.
The group involves Britain, Canada, France, Germany, Italy, Japan and the United States plus the EU.