On the 18th local time, the U.S. daily The New York Times reported that state-backed forces from Iran and China, as well as Israeli companies, have deployed AI-driven autonomous public opinion manipulation on social media.

According to U.S. officials and security experts cited by The New York Times, these nations and entities have rapidly expanded online opinion manipulation activities using AI at speeds and scales surpassing humans, requiring only minimal human intervention.

They utilized freely downloadable and modifiable open-source AI models from China to create autonomous agents capable of performing specific tasks on their own.

Hundreds of AI agents autonomously created fake accounts on platforms such as Instagram, Facebook, X, and TikTok.

They then rapidly disseminated false posts regarding political and current affairs issues through these fake accounts in order to sway public opinion.

It was tracked that the campaigns by China and Iran were state-sponsored, while the Israeli campaign was led by private enterprises.

In the case of the Iranian campaign, it targeted the American public, and the agent-generated fake accounts disguised themselves as ordinary Americans residing in major cities.

These accounts tagged journalists and politicians, circulating popular memes and anti-Republican opinions.

They accumulated approximately 80,000 followers through activities in the first half of this year.

Previously, Meta, which operates Facebook and Instagram, also announced late last month that it had blocked an Iranian propaganda operation utilizing AI.

Officials noted that this marks virtually the first instance where AI agents have taken charge of every step of opinion manipulation activities, ranging from account creation to the coordination of online messaging.

While these AI agent-based activities were rudimentary, they drew intense attention from U.S. intelligence agencies, tech companies, and security experts.

This is because it has sparked fears that publicly available AI models could be abused to carry out online opinion manipulation virtually on their own.

Some U.S. officials expressed concern that similar attempts could be made targeting American voters ahead of the midterm elections in November.

Kyle Crichton, a researcher at Georgetown University's Center for Security and Emerging Technology, predicted that as technology continues to advance, AI agent-led opinion manipulation attempts will become more common.

Researcher Crichton explained that agents can become quite sophisticated, moving beyond simple scale expansion to gain the ability to blend in naturally like real human users.

(Photo: AP, Yonhap News)