The government is expected to consider new safeguards to prevent copyrighted works from being improperly used in generative artificial intelligence services that summarize online news articles and other content to provide answers.
An expert panel under the Intellectual Property Strategy Headquarters on May 25 approved, in principle, the draft of a new Intellectual Property Strategic Program that includes measures to address generative AI.
The draft states that the government will “take stock of issues†concerning generative AI summarization services to “prevent the improper use of copyrighted works.â€
It will be adopted at a headquarters meeting as early as June.
In recent years, services based on “retrieval-augmented generation,†which combine generative AI with search engines, have spread rapidly.
In addition to utilizing data learned previously, these services search for the latest information online and present shortened answers.
They are said to be able to provide more accurate and up-to-date information.
However, rights holders who created the original content, such as newspaper publishers, fear that users may be satisfied with AI-derived responses, reducing visits to their own websites.
There are also concerns about negative impacts on sales of article databases in addition to individual paywalled articles.
The Cabinet Office, the Agency for Cultural Affairs and the Fair Trade Commission will examine solutions to prevent improper use and organize the issues involved as a short- to midterm policy initiative.
For search engines, Cabinet orders and ministerial ordinances already require AI operators to respect the intentions of rights holders when they refuse the use of their content.
The focus of the discussions is whether to amend the framework to establish similar regulations for generative AI.Â
Rights holders dictate what content is off limits to web crawler bots, or software applications, used by AI operators in a “robots.txt†file on their websites.
However, unknown or even disguised crawler names are among the challenges that have been pointed out.
The Intellectual Property Strategy Headquarters is considering introducing a set of rules without penalties for generative AI operators.
While assessing its effectiveness, the headquarters plans to explore additional safeguards.
“It is desirable for AI operators and rights holders to conclude contracts and build a win-win relationship,†said Tatsuhiro Ueno, a professor at Waseda University specializing in copyright law. “If new legal interpretations or institutional reforms move forward, the bargaining power of rights holders could strengthen.â€
The Japan Newspaper Publishers and Editors Association issued a statement on generative AI summarization services in April.
Pointing out a dramatic increase in the risks of free-riding on news content and copyright infringement, the association called for a legal obligation that would compel AI operators to respect rights holders’ “opt-out†decisions.
The Asahi Shimbun and Nikkei Inc. have filed a lawsuit against U.S. startup Perplexity, alleging that the company’s generative AI search service used articles without permission and violated their copyrights. They are seeking damages and other remedies.
