Japanese police to use AI to search social media posts for potential terrorists

Officer AI is expected to patrol feeds next year.

It goes without saying that politics tends to bring out passionate views in some people, and among them are those who would take things to violently extreme lengths. Even though Japan might seem relatively tame in its political rhetoric, it has certainly seen violence with the assassination of former Prime Minister Shinzo Abe and attempted assassination of then Prime Minister Fumio Kishida.

Faced with these very real risks, the National Police Agency has decided to embark on a pilot project next year in which AI will be used to seek out individuals likely to commit terrorist acts by viewing their social media posts.

Of course, the police are already scanning social media for posts. For example, last July, before former PM Kishida was set to give a stump speech at JR Funabashi Station in Chiba Prefecture, they spotted a post that read, “I’ll kill him if he comes.” Following up on it, the police gave the poster a warning and were told that he was just drunk when he did it.

▼ I have to assume being drunk didn’t completely absolve him of making death threats, but at least it lowered the clear and present danger to Kishida at the time.

According to the NPA, that was just one of some 900 similar posts in the month before this year’s election. Having humans find and follow up on all of those threats can be very resource draining, so it’s hoped that AI can ease some of the burden. The proposed plan will have AI find posts from individuals that threaten violence or praise terrorism and then scour that person’s past posts to see if there is a pattern of behavior or if they were just really drunk.

Depending on previous posts, the AI will assign a risk value to that person, so the human police will know which cases to prioritize. The AI will be trained to seek out straightforward words like “bakudan” (bomb) as well as ever-evolving online slang terms like “56su” (kill) and “4ne” (die).

Although the system is intended to streamline procedures that are already in place in Japan, many comments from trending news websites did not welcome AI sifting through their social media accounts.

“They’re going to have to arrest a lot of us.”
“Welcome to censorship.”
“First this, then the Sibyl System.”
“Are we in Minority Report?”
“This all sounds horribly dystopian.”
“I kind of want to know my risk assessment.”
“They’re always watching us.”
“I don’t like an AI looking into my love of Kirby and VTubers.”
“Great, now AI is taking our freedom of expression.”
“Would a real terrorist broadcast what they’re going to do?”

The National Police Agency pointed out that the program was specifically intended to target lone offenders, who are not affiliated with any terrorist organization. Even though we don’t often think about Abe’s shooter as a “terrorist,” his actions perfectly fit the dictionary definition of “terrorism,” and he is probably exactly the type of person they are trying to stop in the future.

Whether AI will be effective at that remains to be seen. I can’t even trust it to give me accurate movie trivia, so I’m not sure I trust it to give me an accurate psychological profile.

Source: Asahi Shimbun, My Game News Flash
Featured image: Pakutaso
Insert images: Pakutaso
● Want to hear about SoraNews24’s latest articles as soon as they’re published? Follow us on Facebook and Twitter!

Credit: