AI allegedly ‘Piggybacking’ off Media Organizations’ Work; Information Taken from Articles Requiring Paid Membership

The Yomiuri Shimbun

20:00 JST, October 17, 2023

The Japan Newspaper Publishers and Editors Association provided concrete examples of the unauthorized use of articles owned by media organizations by generative artificial intelligence models to emphasize the necessity of the protection of copyrights, at a subcommittee of the　Cultural Council on Monday.

Amid mounting risks of the spread of false information by generative AI, the government and ruling parties are seeking to make use of Originator Profile (OP), a digital technology which identifies senders of information, on a global basis.

At the subcommittee, the association exemplified a case of suspected unauthorized use of a media organization’s article using U.S. tech giant Microsoft’s conversational AI-powered search engine, Bing AI.

According to the association, when a user typed in “Please tell me about the ‘kakusei’ guest star that suddenly appeared in the night sky in southern Japan more than 1,000 years ago, during the Heian period,” the outputted response matched the text of an article published online by The Yomiuri Shimbun on Aug. 10 in multiple places. The match rate ranged from 35% up to 52% for each of the five paragraphs.

Other examples were also presented, including one in which a response was created based on an article limited to paid members of The Mainichi Shimbun’s digital service using Google’s SGE search service, which is equipped with generative AI.

Article 30-4 of the Copyright Law, established in a 2018 revision, allows AI to learn copyrighted works without permission. The association said, “As long as Article 30-4 exists, the piggybacking of news content [by generative AI] is inevitable,” and underlined the necessity of a revision to the law.

The development of digital technology has led to the spread of information that is not based on accurate reporting or facts.

Over the clashes between Israel and the Islamist organization Hamas, a fake video spread on X (formerly known as Twitter) claiming that Ukraine is selling weapons provided by NATO to Hamas. It bears the logo of the BBC, but BBC reporters and others have denied that they ever disseminated such information.

If the use of generative AI that produces elaborate text and images without specifying sources becomes widespread, it could spur social disorder.

OP is expected to be a means of maintaining the integrity of the internet space. It is a technology that can electronically assign third-party-authenticated information about the senders to articles and other materials on the internet. The Originator Profile Collaborative Innovation Partnership, in which 31 corporations including The Yomiuri Shimbun participate, aims to put this technology to practical use in 2025. On Oct. 3, NHK also announced its participation in the partnership.

In an interview with The Yomiuri Shimbun on Oct. 2, Prime Minister Fumio Kishida praised OP as “contributing to countermeasures against the spread of disinformation.”

Also regarding the “Hiroshima AI Process,” in which the Group of Seven nations will discuss how to regulate generated AI, Kishida stated, “OP is worth taking up as one of the items for cooperation.”

On Oct. 11, coalition partner Komeito’s headquarters for the promotion of digital society also submitted a proposal to the Prime Minister’s Office. They presented OP as “an effective technology for combating disinformation, copyright infringement and increasingly sophisticated crimes,” and urged the government to take the lead in making the OP technology an international standard.