• Social Series

Unbalanced information diet: AI-generated deception / OSINT Reveals Discrepancies in Gaza Hospital Explosion Images; Technology Developed to Recognize AI-Generated Material

The Yomiuri Shimbun
Eliot Higgins speaks during an online interview with The Yomiuri Shimbun on Nov .1.

This is the final installment in a series examining moves to spread biased information through the use of generative artificial intelligence, acts that threaten our democracy.

***

Eliot Higgins, founder of the private research organization Bellingcat, has popularized the method of “Open Source Intelligence” (OSINT) worldwide, which investigates and analyzes public information to find the truth.

The Middle Eastern satellite television network Al Jazeera reported in a breaking news story on Oct. 17 that 500 people had been killed in an explosion at Al-Ahli Hospital in Gaza.

The Islamist group Hamas claimed the hospital was attacked in an Israeli strike. Israel countered that the hospital was hit by a missile launched by a Hamas-affiliated group in Gaza. It was debatable which side actually launched the attack.

Bellingcat quickly analyzed numerous videos and images of the aftermath posted on social media.

The following day, Bellingcat published a report on the investigation under the title: “Identifying Possible Crater from Gaza Hospital Blast.”

Bellingcat indicated that the hospital blast was not caused by the Israeli military, stating in a report that the size of the crater was too small to be caused by typical Israeli munitions.

Human Rights Watch, an international human rights organization, also conducted its own investigation and indicated that the explosion may have been a misfire by a Palestinian armed group.

Higgins said “Because this is the worst situation in which a large amount of disinformation is circulating, Bellingcat immediately disproved it and told only the facts for what the evidence can tell.”

Concerns over AI evolution

There are researchers in Japan who analyze satellite and social media images, such as Yu Koizumi, associate professor at the University of Tokyo’s Research Center for Advanced Science and Technology, who is an expert on Russian affairs.

He showed us a single satellite image, taken of a Russian military base on the northern territories’ Etorofu Island. Before the Russian aggression into Ukraine began in February 2022, the base was equipped with S-300 surface-to-air missiles. However, they have not been seen since autumn that year.

Antiaircraft missiles that were deployed in other regions in the Far East have also disappeared one after another. Koizumi believes that these missiles were probably redeployed to the Ukrainian front.

In February last year, the Russian military announced that it had conducted S-300 launch drills in the vicinity of the island of Etorofu.

The Yomiuri Shimbun
University of Tokyo Associate Professor Yu Koizumi speaks at the Research Center for Advanced Science and Technology in Tokyo on Nov. 27.

“Russia might have released false information in order to show that it is maintaining its strength in the Far East as well,” Koizumi said.

Koizumi also expressed concerns about the evolution of generative AI. “If a large number of false images are circulated, OSINT’s methods could collapse,” he said.

Originator Profile

Originator Profile (OP) is a digital technology that makes it possible to reliably identify articles. A key feature of the technology is that clicking on an online article brings up a frame displaying the newspaper’s name and its reporting guidelines.

The sender’s information embedded in the article does not disappear even if it is copied to asocial media.

Shigeya Suzuki, a specially-appointed professor at Keio University who leads the development division of the OP Collaborative Innovation Partnership said that the goal of the first stage of the verification experiment that began in July has been completed.

OP is also said to be effective in countering disinformation generated by AI.

Through the use of OP technology, AI-generated information will be tagged so that recipients can recognize the information as being “AI-generated”.

The Yomiuri Shimbun
Shigeya Suzuki, a specially-appointed professor at Keio University, left, explains about an experiment of the Originator Profile in Minato Ward, Tokyo, on Nov. 24.

There are concerns that the quality of AI-generated information will deteriorate rapidly as AI relearns information that already contains AI-generated errors.

However, this tag would prevent such relearning, they said.

The partnership aims to standardize OP internationally. “OP is highly flexible and has a wider range of applications than expected,” Suzuki said.

The Content Authenticity Initiative (CAI), which has more than 2,000 member companies worldwide, including news organizations, is also promoting similar initiatives in the area of images.

The CAI intends to embed information into the image, such as who took the photo and how it was processed, so that it can be verified even if it has been processed by AI.

“Creating a system that allows people to check the ‘history’ of information on the internet is important for society as a whole,” said Shoichi Nishiyama, chief digital officer at the Japan arm of U.S. software giant Adobe Inc., which operates the CAI office.

There is no special medicine to protect society from information that manipulates the public will. It is necessary to implement all kinds of countermeasures, including OSINT, OPs and information literacy.