JAPAN'S DRAFT AI GUIDELINES CALL FOR CURBING DEEPFAKES, BUILDING LITERACY

TOKYO, Nov 12 (Bernama-Kyodo) -- Japan's new guidelines for the appropriate use of artificial intelligence will urge companies to curb "deepfakes" and other misleading outputs, while encouraging users to build AI literacy, a draft outline showed Wednesday, Kyodo News agency reported.

The guidelines, which would not be legally binding, also call on operators to ensure transparency by disclosing necessary information, as the government seeks to ease public concern over generative AI, according to the document.

Based on an AI law that took full effect in September, the government is drafting the guidelines along with a broader basic policy to promote wider use of AI.

The proposed guidelines outline key points for both users and businesses to keep in mind when using the technology.

Specifically, the draft calls on developers and operators to establish and disclose policies regarding the collection of training data to mitigate the risk of biased outputs and privacy violations.

It also emphasises the need to use the latest technology and expertise to tackle the issue of "hallucinations," in which inaccurate information is presented as fact.

Users, meanwhile, are urged to be mindful of potential risks associated with AI, such as bias reinforcement and criminal misuse, according to the draft.

Central and local governments that use AI must "fulfil accountability to citizens" by ensuring that the basis for administrative decisions remains transparent, the document said.

The guidelines are expected to require each ministry and municipality to appoint an official responsible for overseeing AI-related risks.

-- BERNAMA-KYODO