Japanese police have decided to use artificial intelligence to step up cyber patrols to detect online information on homemade guns and explosives.

The National Police Agency plans to introduce in September a technology called natural language processing, which analyzes texts and singles out content that could pose serious risks to human life.

The agency has been asking website managers to delete content on how to make firearms and explosives since February.

The technology will be applied to posts on Twitter and YouTube. The agency expects that AI will detect harmful content more accurately than manually entering keywords and searching.

The move toward greater online scrutiny comes after former Prime Minister Abe Shinzo was fatally shot by a man carrying a homemade gun last July.

In another incident, on Saturday, a man was arrested after he threw what appeared to be a homemade explosive device towards Prime Minister Kishida Fumio. He was unharmed.