With generative artificial intelligence gaining rapid adoption worldwide, private-sector businesses in Japan are starting to embrace related information search technologies to enhance operational efficiency.

Conventional information searches require users to enter relevant keywords and manually browse selected websites to locate the desired information. The process can be time-consuming, however, and users may not always find the exact content they need.

Generative AI search services are offered by companies such as Google, OpenAI, and the U.S. startup Perplexity.

The technology is also proving valuable for searching corporate databases. Lion, a leading Japanese consumer goods company, has developed a generative AI-powered search system that allows its researchers to efficiently search through research reports and conference materials stored in the company’s database.

Lion’s search system uses AI to process user queries and provide text-based answers derived from multiple documents. It also displays the original documents on-screen, enabling users to review the source material instantly.

Shintaro Yamaoka, manager of the company’s digital strategy department and a key figure in the system’s development, said, “To drive innovation, we need a system that allows researchers to have quick access to technical knowledge when they need it.” According to Yamaoka, the time required to locate a desired document has been reduced to an average of one-fifth of that with conventional search systems.

The AI-based system’s translation function enables researchers from abroad to instantly understand research reports in their native languages, enhancing the system’s positive reception among employees.

Yuki Yuri from the department said the system is “beneficial not only for new recruits and young employees but also for those transitioning to new workplaces, as it helps them systematically acquire knowledge in unfamiliar fields.”

Still, generative AI searches face numerous challenges.

One significant issue involves copyright concerns for services that utilize information from the internet. In a statement issued in July, the Japan Newspaper Publishers & Editors Association noted that content from news organizations is sometimes used without permission in generative AI searches, which likely constitutes copyright infringement. The statement called for a review and improvement of relevant laws.

Additionally, generative AI is prone to “hallucinations,” or generating information that contradicts established facts. While combining it with traditional information searches can help reduce this risk, further countermeasures are still necessary.

During a June interview in Tokyo, Perplexity CEO Aravind Srinivas explained the company’s responses to the problem.

“The advantage of a service like Perplexity is it doesn’t use one source to answer your question. It uses many sources. … What the AI here can do is provide the user perspectives from multiple sources,” he said.

“At least we’ll try to source human content, not AI content” to answer user queries, emphasizing that this approach helps “prevent misinformation as much as possible,” he added.

The convenience of generative AI searches makes it all the more important for regulators and lawmakers to swiftly develop a robust legal framework, for service providers to implement high-quality measures, and for users to cultivate high standards of AI literacy.