October 19, 2023
TOKYO – The G7 has agreed in principle on international guidelines for developers of generative AI, according to sources.
The draft of the agreement has 11 items, including items concerning adequate copyright protection and calls to take appropriate measures to reduce risks at the development stage, the sources said. The AI developers in mind include companies and research institutes.
An official agreement on the guidelines will be sought during a virtual G7 summit to be held as early as this autumn.
Basic agreement was reached during an informal meeting of the Hiroshima AI Process, a framework for the G7 to discuss how to use and regulate generative AI, held on Oct. 9 on the sidelines of the U.N.-sponsored Internet Governance Forum in Kyoto.
Junji Suzuki, internal affairs and communications minister, represented Japan during the generative AI talks, which followed on the heels of September’s ministerial-level meeting of the Hiroshima AI Process.
According to the draft agreement, AI developers will be requested to identify risks and security weaknesses at the development stage and take appropriate measures. Protection of personal information and copyrights are required when AI is learning from data.
As for measures against disinformation, the draft calls for the development of digital watermarking technology that can differentiate documents and images created by AI, as well as technologies to prove content is reliable. This apparently refers to Originator Profile digital technology for indicating the source of online information.
The G7 is also discussing guidelines for providers and users of AI-based services as part of the Hiroshima AI Process. Another ministerial-level meeting, perhaps as soon as December, will finalize the plans and prepare reports to the leaders of the G7.