China is utilizing faux social media accounts to ballot voters on what divides them most to sow division and presumably affect the end result of the U.S. presidential election in its favor. China has additionally elevated its use of AI-generated content material to additional its targets all over the world. North Korea has elevated its cryptocurrency heists and provide chain assaults to fund and additional its navy targets and intelligence assortment. It has additionally begun to make use of AI to make its operations simpler and environment friendly.
These are among the many Microsoft Menace Intelligence insights within the newest East Asia report printed right now by the Microsoft Menace Evaluation Heart (MTAC) – Similar targets, new playbooks: East Asia menace actors make use of distinctive strategies.
Three key findings emerge within the report:
- Misleading social media accounts by Chinese language Communist Social gathering (CCP)-affiliated actors have began to pose contentious questions on controversial U.S. home points to raised perceive the important thing points that divide U.S. voters. This may very well be to collect intelligence and precision on key voting demographics forward of the U.S. presidential election.
- There was an elevated use of Chinese language AI-generated content material in current months, trying to affect and sow division within the U.S. and elsewhere on a spread of subjects together with: the prepare derailment in Kentucky in November 2023, the Maui wildfires in August 2023, the disposal of Japanese nuclear wastewater, drug use within the U.S. in addition to immigration insurance policies and racial tensions within the nation. There’s little proof these efforts have been profitable in swaying opinion.
- China’s geopolitical priorities stay unchanged however it has doubled down on its targets and elevated the sophistication of its affect operations (IO) assaults. These priorities are:
- The South Pacific islands
- The South China Sea area
- The U.S. protection industrial base
Inauthentic Chinese language social media accounts attempt to study extra on what divides US voters
MTAC beforehand reported in September 2023 how CCP-affiliated social media accounts had begun to impersonate U.S. voters in an effort to affect the 2022 U.S. midterm elections. This exercise has continued and these accounts almost solely publish about divisive U.S. home points corresponding to international warming, U.S. border insurance policies, drug use, immigration, and racial tensions. They use unique movies, memes, and infographics in addition to recycled content material from different high-profile political accounts. In current months, there was a rise in, successfully, polling questions. This means a deliberate effort to grasp higher which U.S. voter demographic helps what problem or place and which subjects are essentially the most divisive, forward of the primary part of the U.S. presidential election.
China will increase use of AI in affect campaigns
Chinese language IO operations within the U.S. continued to opportunistically bounce on occasions which might serve their strategic pursuits – corresponding to portraying the U.S. in an unfavorable gentle. These operations, attributed to Storm 1376, included:
- Urging audiences to contemplate whether or not the derailment of a prepare carrying molten sulfur in Kentucky in November 2023 was intentionally attributable to the U.S. authorities and whether or not it’s “intentionally hiding one thing”. Some messages even likened the derailment to 9/11 and Pearl Harbor cover-up theories.
- Accusing the U.S. of purposefully poisoning different nations’ water provides to keep up “water hegemony”. This was a part of a wider multilingual marketing campaign, principally targeted on Japan and its authorities’s determination to get rid of handled radioactive wastewater into the Pacific Ocean. Storm-1376 tried to solid doubt on the Worldwide Atomic Power Company’s (IAEA) scientific evaluation that the disposal was secure.
The Taiwanese presidential election in January 2024 noticed a surge in using AI-generated content material to enhance IO operations by CCP-affiliated actors. This was the primary time that Microsoft Menace Intelligence has witnessed a nation-state actor utilizing AI content material in makes an attempt to affect a overseas election.
The group we name Storm-1376, also called Spamouflage and Dragonbridge, was essentially the most prolific. For instance, on election day, it posted suspected AI-generated faux audio of Foxconn proprietor and election candidate Terry Gou (who had bowed out of the competition in November 2023) endorsing one other candidate within the presidential race. Gou had made no such assertion. YouTube shortly eliminated this content material earlier than it reached a wider viewers.
Storm-1376 has promoted a sequence of AI-generated memes of Taiwan’s then-Democratic Progressive Social gathering (DPP) presidential candidate William Lai, and different Taiwanese officers in addition to Chinese language dissidents all over the world. These have included an rising use of AI-generated TV information anchors that Storm-1376 has deployed since at the very least February 2023.
North Korean cyber operations
North Korea continued to prioritize the theft of cryptocurrency funds, conducting software program supply-chain assaults and concentrating on their perceived nationwide safety adversaries. That is prone to generate income, principally for its weapons program, along with accumulating intelligence on the US, South Korea, and Japan.
The United Nations estimates that North Korean cyber actors have stolen over $3 billion in cryptocurrency since 2017. Heists totaling between $600 million and $1 billion occurred in 2023 alone.
Our report catalogs a number of cases of cryptocurrency heists, spear-phishing, and software program supply-chain assaults and efforts to undermine the trilateral alliance between the U.S., Japan, and South Korea.
Notably, Microsoft and OpenAI have noticed the North Korean actor we name Emerald Sleet utilizing instruments powered by AI large-language fashions (LLMs) to make their operations simpler and environment friendly. Microsoft partnered with OpenAI to disable accounts and belongings related to Emerald Sleet.
Wanting ahead
With main elections happening all over the world this 12 months, notably in India, South Korea and the US, we assess that China will, at a minimal, create and amplify AI-generated content material to learn its pursuits. Regardless of the probabilities of such content material in affecting election outcomes remaining low, China’s rising experimentation in augmenting memes, movies, and audio will probably proceed – and should show simpler down the road. We are able to anticipate to see North Korea proceed to steal cryptocurrency to fund house, missile, and nuclear packages in addition to launch supply-chain assaults on the protection sector.
