AI Regulation: Fake Videos Circulating Online More Serious Than Just Pranks

It can be said that the “negative aspect” of generative artificial intelligence, which enables anyone to easily create fake videos and images, has been exposed. The government should not underestimate the adverse effects of fake videos.

A fake video of what appears to be Prime Minister Fumio Kishida speaking to the public at an online press conference has circulated on social media. A man in his 20s from Osaka Prefecture admitted in an interview that it was a fake video created with generative AI.

In explaining the reason for creating the fake video, the man said it was “something like satire,” but a fake video could cause confusion in society depending on the content, and it must be said that what the man did was nothing but malicious.

The fake video was created by altering a video clip of an actual online press conference held by the prime minister, and the man used footage from a Nippon TV news program without permission. To make the video look more authentic, it included the program’s logo and the word “LIVE” to indicate that it was a live broadcast.

The person seen in the video was taken from footage of the real prime minister, but the audio was created by AI. The man said he had AI learn from audio of the prime minister’s speeches and press conferences, which are widely available online. The man converted his own voice into a fake voice that sounded just like Kishida.

Although few people may have taken the prime minister’s “remarks” seriously because of the vulgar content, the video was viewed several million times. Some people may have been baffled, wondering, “What on earth is this?”

This is not the only case of the abuse of AI. A university student in Aichi Prefecture posted a fake image of a female athlete in her 20s who belongs to a university’s athletics team on social media. The image was a composite of her face and another person’s naked body. The Metropolitan Police Department has sent documents pertaining to the man to prosecutors on suspicion of defamation in violation of the Penal Code.

The man told the police that he wanted to “satisfy his desire,” but if found guilty of defamation, he will likely have to pay a heavy price: imprisonment for up to three years with or without labor, or a fine of up to ¥500,000.

However, a creator of AI-generated disinformation will not be held criminally responsible if the fake content does not cause actual harm, such as defamation or damage to a company.

In order to prevent the spread of confusion in society, it is essential to prevent a flood of false information. Legal restriction of the creation of false information itself is an important issue to consider.

One of the reasons behind the abuse of generative AI is the 2018 revision of the Copyright Law, which in principle allows AI to learn from copyrighted works without the permission of the rights holders.

Regulations on AI that are more lax compared to those in other countries may have reduced a reluctance to use AI in modifying news articles and programs that are copyrighted works. The government needs to hasten further revision of the Copyright Law.

(The Yomiuri Shimbun, Nov. 7, 2023)