AI use rising in influence campaigns online, but impact limited - US cyber firm

News By Daniel Michan Published on August 22, 2023

According to Mandiant, a cybersecurity firm owned by Google, there has been an increasing trend in using intelligence (AI) for conducting manipulative online campaigns. While the application of AI in digital intrusions has been limited so far, Mandiant's researchers have observed numerous instances since 2019 where AI-generated content, including fake profile pictures has been employed in politically motivated influence campaigns.

These campaigns have been associated with governments like Russia, China, Iran, Ethiopia, Indonesia, Cuba, Argentina, Mexico, Ecuador, and El Salvador. The report also highlights the surge in generative AI models like ChatGPT that facilitate the creation of convincing fake videos, images, texts and computer code. Security officials have expressed concerns about cybercriminals exploiting these models.

Mandiant researchers suggest that generative AI could empower groups with resources to produce high-quality content for influence campaigns on a large scale. For example, the Dragonbridge information campaign supporting China has experienced growth across 30 social platforms and ten languages since it initially targeted pro-democracy protesters in Hong Kong in 2019. Sandra Joyce, Vice President at Mandiant Intelligence, cited this as an expansion.

However, the impact of campaigns could have been more extensive. According to her, they have yet to succeed in altering the current threat landscape.

China has consistently denied involvement in influence campaigns as accused by the United States.

Mandiant, a company that assists private organizations in responding to digital breaches, stated that they have not observed AI playing a significant role in threats from Russia, Iran, China, or North Korea. The researchers expect the use of AI for intrusions to remain minimal in the near future.

"To this point, we haven't encountered any incident response where AI has played a role," Joyce mentioned. "We haven't seen it being practically utilized more effectively than conventional tools."

Nonetheless, she also emphasized that this problem is expected to grow over time.