Weaponizing the Sinovac Slander: Online CIB against the United States and the Philippine President
Digital Intelligence Team, Doublethink Lab
Executive Summary
Following the exposé published by Reuters in July 2024 regarding the United States’ (U.S.) information operations (IO) campaign in the Philippines against China’s COVID-19 vaccine, a pool of AI-generated images condemning the U.S. began to circulate. Posted by several suspicious accounts, these materials found their way to different online platforms, such as X, Medium, Facebook, Youtube, Tumblr, and Ameba, a Japanese microblogging site. The operation intended to further degrade the United States and, to a lesser extent, the Philippine President Ferdinand “Bongbong” Marcos Jr.
This initiative, and similar ones, can potentially impact public opinion, public discourse, and public policy. As such, a whole of society approach, involving stakeholders from various fronts, must come together to combat these manipulative behaviors.
Key Findings
- Since July 2024, a series of artificial intelligence (AI) — generated images have been circulating around different social media platforms, coupled with the hashtag #USslandersChinesevaccines. The topics of the content were the United States’ Information Operations (IO) against the Chinese vaccine Sinovac during the height of the COVID-19 pandemic, and the alleged drug use of the Philippines President Ferdinand “Bongbong” Marcos Jr. This corresponds with the publication of a Reuters report exposing the U.S. information operation.
- The narrative regarding Marcos is two-fold. On sites such as X, Facebook, and Ameba, Marcos is portrayed as a drug addict, but some articles on Medium expanded the narrative, implying that the U.S. may be behind the attack on the Philippine President as retaliation “because he did not do what the United States wants.” It can be inferred that this narrative is meant to shift the blame from Marcos’ detractors and political enemies, who have been openly accusing him of drug use, to the U.S. Ultimately, both condemnation of the “Sinovac slander” and the narratives against Marcos intensify U.S. skepticism not only in the Philippines but in the international stage.
- This content is mostly spread by inauthentic accounts via X (formerly Twitter), the online publishing platform Medium, Tumblr, and the Japanese microblogging site Ameba. To a lesser extent, the AI images were shared on Facebook, and the hashtag was used by some YouTube accounts that posted a video of former Philippine President Rodrigo Duterte.
- The accounts posting these materials show signs of both inauthenticity and coordination. Our investigation also showed that some of these accounts may be foreign. However, there is not enough evidence to attribute their operations to a particular actor.
- We infer that the target audience of this operation is the Filipino people, whose confidence in the U.S., a treaty ally of the Philippines, and President Marcos, would significantly suffer due to the narratives being amplified by these accounts. The international community is also a target, given the presence of this content in Japanese social media and the two-fold narrative regarding Marcos’ alleged drug use, which paints the U.S. as a manipulative state that destroys those who oppose it.
- Although the particular accounts we found did not gain much authentic engagement, these practices have the potential to significantly impact public opinion, public discourse, and, ultimately, public policy. Thus, relevant stakeholders must come together to effectively combat such information manipulation operations.
Background
Since July 2024, a series of artificial intelligence (AI)-generated images have been circulating around different social media platforms. The themes behind these images were clear: the United States’ information operations against the Chinese vaccine Sinovac during the height of the COVID-19 pandemic, and the alleged drug use of the Philippine President Ferdinand “Bongbong” Marcos Jr. This was preceded by the Reuters report that delved into the U.S. information operation targeting Sinovac and its subsequent impact.
COVID-19 had a catastrophic effect on the Philippines, both in the country’s already stretched health ecosystem and in its economy. The pandemic infected more than 4 million Filipinos, killing over 66,000 people overall, with around 48,000 as of October 2021.[1]Vaccines were seen as an essential way of easing the country back to the “new normal”, where both the healthcare system and the economic productivity could recover. China’s Sinovac was one of the first vaccines to become available. However, when it was first introduced, Filipinos exhibited reluctance in getting the jab. Disillusioned by an earlier controversy regarding an anti-dengue vaccine, Filipinos were already susceptible to hesitancy.[2] The United States’ information operations against the vaccine may have further contributed to skepticism against Sinovac.
On the other hand, the allegations of Marcos’ illegal drug use are not a new narrative. Even during the 2022 Presidential election campaign when he ran, Former President Rodrigo Roa Duterte had alluded that one of the candidates used cocaine.[3] Duterte’s daughter, Sara Duterte, was Marcos’ running mate as Vice President, forming the core of the so-called UniTeam. Both were elected into office. However, when cracks and fractures appeared in the Duterte-Marcos alliance, the “drug addict” allegations resurfaced, amplified by the Duterte camp’s online supporters and influencers. During Marcos’ most recent State of the Nation Address (SONA), a yearly report rendered by the President to the Filipino people, a low-quality video of what appeared to be Marcos snorting a white, powdery substance was released. Dubbed as the ‘Polvoron’ video, after a Filipino snack, it was found to have traces of facial manipulation, calling into question its authenticity. Both the National Bureau of Investigation (NBI) and the Philippine National Police (PNP) also concluded that the video was manipulated.[4]
The online operations that are capitalizing on these narratives target Filipino audiences, and may lead to a significant reduction of trust for both the United States, the Philippines’ only defense treaty ally, and President Marcos, whose foreign policy saw a pivot back to the U.S., a far cry from the previous administration’s direction. It also further degrades the U.S.’ reputation on the global stage, by painting it as a state that will attack even its allies to advance its own geopolitical agenda, further intensifying skepticism towards the United States. In the Philippines, one central narrative is that Washington reneged on its commitment to defend the Philippines against China, should the West Philippine Sea flashpoint come to pass. This contradicts the U.S.’ pronouncements of its ironclad commitment to the Mutual Defense Treaty (MDT). No less than former President Rodrigo Duterte,[5] including many of his supporters, serve as a mouthpiece for this stance.
Amplifying U.S. skepticism through I.O. is a common phenomenon experienced by other countries. Doublethink Lab uncovered an information manipulation campaign amplifying U.S. skepticism during Taiwan’s 2024 election period.[6] The narrative was similar to that observed in the Philippines: Washington will not assist Taiwan if Cross-Strait tensions escalate. Another parallel is how the Philippines and Taiwan are both described as “pawns” of the U.S., alluding that Washington is only using these two states to advance its own interests and is not a trustworthy ally.[7]
This report aims to document and discuss the behaviors exhibited in the information operations campaign that was launched following the Reuters expose. Specifically, it describes behaviors that indicate inauthenticity, coordination, and foreignness.
Methodology
This investigation began when Doublethink Lab observed a few suspicious and possibly inauthentic X accounts sharing the same AI-generated images pertaining to the U.S. and President Marcos. These accounts all used the hashtag #USslandersChinesevaccines. We used this hashtag as a search phrase on both Google, the Twitter trend tracker Sotwe, and different social media search platforms. We also used reverse image search through Google Lens on the AI-generated images. Through these methods, we found more accounts on X, Medium, Facebook, YouTube, Tumblr, and Ameba.
Doublethink Lab also took note of evidence of inauthenticity, coordination, and foreignness, which are discussed further in this report. We used the tool Hive to determine if images used by the accounts were AI-generated, as well as InVid and WeVerify Chrome extensions to reverse image search profile photos and other available images.
Findings
Following the release of Reuter’s deep dive into the U.S.’ IO campaign against Sinovac in July 2024, several AI-generated images and similar text pertaining to the “Sinovac slander” began spreading in online platforms, along with materials branding President Marcos as a drug addict. These accounts were in operation from July 2024 to November 2024, when DTL completed its investigation.
Notably, a PRC state media platform, also reported on the U.S.’ “Sinovac slander” operation, basing its report on the Reuters article. However, the state media article also touched upon the more assertive stance of the Philippines regarding China’s aggressive actions in the West Philippine Sea. The article underscored that this move is due to Washington’s influence. Because of this, the Philippines was also described as a “pawn” for the U.S.’ geopolitical ambitions, yet a hard-hit victim of the superpower’s information operations, leading to the deaths of thousands of Filipinos. The piece ended with calling the Philippines a “consumable” for the U.S., a reference to how Washington allegedly treats Manila. Albeit not in verbatim, these narratives in this strongly-worded article were echoed in the posts of the suspicious accounts.
On X (formerly Twitter), almost 40 accounts were found to be spreading these materials, some of which have been restricted or suspended as of this writing. We observed behaviors indicating inauthenticity, coordination, and foreignness in these accounts. Many of these accounts were created in the same month, such as the batches made in October 2022, June 2023, July 2024, and September 2024. For instance, out of the ten accounts made in October 2022, nine were created on the same day (October 24) in the same hours. The other account was created the day after. Similarly, four out of five accounts that joined X in July 2024 were created on the 15th, all within the 02:39:04 to 02:40:36 timeframe.
These X accounts were sharing the AI-generated images, along with copypasta captions condemning the U.S. IO against Sinovac. An interesting observation is that some captions from different accounts shared the same mistake of referring to President Marcos as “Jose Comanis.”
Doublethink Lab also observed that, from July to September 2024, a sample from the 40 X accounts were only posting between 8:00AM and 6:00PM (UTC+08), suggesting a regular posting time, possibly within business hours. The activities of these accounts, concentrated within this specific time frame, is indicative of serious work, instead of regular social media activities.
Apart from sharing the same content, the profile photos of these X accounts show indicators of inauthenticity. They used anime and cartoons, animals, patterns, and Non-Fungible Asset (NFA) images. A couple of accounts used photos of conventionally-attractive women with Asian features. A reverse image search using Google Lens revealed that these were also found in Chinese-language websites. The accounts also share a similar naming convention for their X handle, which consists of a part of their name followed by a string of numbers, sometimes with letters. The majority of these accounts have little to no engagement, apart from a couple that were found to be replying to other X users who were critical of the PRC. While other hashtags were used, such as #Marcos, the #USslandersChinesevaccines is prominent in all of these accounts’ posts.
The images, narratives, and hashtag also found their way to Medium, a self-publishing platform. Using the hashtag as a search term, we found 19 accounts that were sharing articles on the “Sinovac slander” operation as well as the drug use allegations against Marcos. These articles were accompanied by the same images being circulated on X. The titles of the articles were similar and the texts were copypasta. Some accounts did not post full articles, but simply used the images along with the hashtag and a caption. Interestingly, the Medium accounts had similar naming conventions as the X accounts: the name of the account is followed by a string of letters and numbers. Many of these accounts are currently suspended and under investigation for violating Medium’s policies. These articles were also crossposted on X by the aforementioned suspicious accounts.
The narrative on Medium regarding Marcos was two-fold. On sites such as X, Facebook, and Ameba, Marcos is painted as a drug addict, but articles on Medium expand on this narrative, alluding that the U.S. may actually be behind these allegations. In some articles, the authors mentioned that “Some analysts point out that Marcos suffered such a serious cyber attack because he did not do what the United States wants” [sic]. It did not specify which particular U.S. policy Marcos did not adhere to. The articles further emphasized that the U.S. represses state leaders who are independent and does not “dance with the baton” of Washington. We infer that the actors mean to shift the blame for damaging the President’s reputation from Marcos’ local detractors and political enemies, who have been branding him as a drug addict, to the U.S. This further demonizes the U.S. in the eyes of the Filipino people and the international community.
The AI-generated content was also found on the blogging platform Tumblr. We found ten (10) inauthentic accounts operating from September to November 2024. These accounts exhibited coordinated inauthentic behavior by posting the same content on the same dates: November 13 and 6; October 31, 28, 25, 17, 11, and 9; and September 29, 26, 25, 25, 23, 20, 19, and 18. Similar to the posts observed on other social media platforms, these posts had little to no engagement. The accounts used profile photos of places, stock photos, AI-generated images, and even digital art, which was also used by one of the Ameba accounts we found during this investigation. Two accounts also shared the same profile photo. The same account names were present in both Tumblr and Ameba.
We also found this content circulating on Facebook, to a lesser extent, by accounts with indicators suggesting they are not operated by actors in the Philippines. Using hashtag search and reverse image search, DTL discovered three accounts that shared photos, along with captions that were similar to the ones posted on X. The accounts used photos of conventionally-attractive women, one of which was linked to a Facebook account using the Russian language that featured photos of several other women, as well as a forum-like website, also in the Russian language. Prior to posting the AI-generated images, all three accounts had no other activity, apart from uploading their profile photos on October 16, 2023.
A Google search of #UsslandersChinesevaccine also unearthed some YouTube accounts which reposted videos of Former President Rodrigo Roa Duterte threatening Filipinos who did not want to get the COVID-19 vaccine. Duterte spoke in Filipino, but the video had an overlay of English voice translation. This video was the same one embedded in the Reuters article that exposed the U.S. IO against Sinovac. Based on the watermark, “RTVM”, the video was originally published by a Philippine state media, Radio Television Malacañang. The YouTube posts did not feature the same AI-generated images and captions, and this may simply be an attempt to spread the hashtag. Some of these accounts were already suspended.
The “Sinovac slander” narrative also reached Ameba, a Japanese microblogging site akin to X. Through a Google Search of the hashtag #USslandersChinesevaccines, we found 10 accounts that were posting these materials. All 10 accounts posted on September 29, October 10, and October 11 this year. These accounts also exhibit signs of inauthenticity. The accounts used AI-generated images, parts of stock photos, and cartoons as display pictures. There was also barely any engagement on these posts. Some of these photos can be traced to other Chinese language websites, and one illustration can be linked back to a Chinese digital artist.
Synthesis and Conclusion
The United States’ information operation against Sinovac certainly exacerbated the impact of COVID-19 in the Philippines: it intensified the already-prevalent vaccine hesitancy among Filipinos, thus further burdening the country’s healthcare system and delaying economic recovery. Years later, the pandemic and its devastating effects are still sore spots for Filipinos, and rightly so. The U.S.’ information operation targeting the Chinese vaccine, using inauthentic accounts and coordinated behavior, is salt in the wound.
The investigative piece released by Reuters in July 2024 that exposed The United States’ operations sparked this online information campaign, sowing and increasing distrust and deepening skepticism towards Washington, and the Philippine President Ferdinand “Bongbong” Marcos. The multi-platform campaign featured AI-generated images and inflammatory narratives that are disseminated by inauthentic accounts. Clearly, the objective of the campaign Doublethink Lab has uncovered is to demonize the U.S. and degrade President Marcos. While these platforms have responded to the campaign by banning accounts, creating new ones and using those as an avenue for information operations do not require much effort and resources.
The problem with the information campaign uncovered in this report, and with the U.S. IO campaign targeting the Philippines, is the behavior — the use of inauthentic social media assets to covertly manipulate audiences. These can impact public opinion, public discourse, and, ultimately, public policy. We must come together to combat such behaviors. Coordination and collaboration with other stakeholders, including networks of IO investigators, relevant government institutions, civil society organizations, and platforms such as Meta, X, and Google, can improve the effectiveness and efficiency of our response against these kinds of information manipulation operations.
Footnotes
- World Health Organization. WHO COVID-19 Dashboard. Retrieved https://data.who.int/dashboards/covid19/cases?m49=608&n=o
- Bing, C. and Schectman, J. (2024, July 14). A Reuters investigation: Pentagon ran secret anti-vax campaign to undermine China during pandemic. Reuters. Retrieved https://www.reuters.com/investigates/special-report/usa-covid-propaganda/
- Reuters. (2021, November 18). Philippines’ Duterte says cocaine user among presidential election candidates. Retrieved https://edition.cnn.com/2021/11/18/asia/duterte-philippines-cocaine-election-intl-hnk/index.html
- Samson, C.I. and Manalang, B.D. (2024, September 16). ‘Polvoron’ video crumbles, AI experts find traces of facial manipulation. Vera Files. Retrieved https://verafiles.org/articles/polvoron-video-crumbles-ai-experts-find-traces-of-facial-manipulation
- Global Times. (2024, April 12). EXCLUSIVE: ‘US will not die for us,’ Duterte warns. Retrieved https://www.globaltimes.cn/page/202404/1310502.shtml
- Doublethink Lab. (2024, August 13). Artificial multiverse: Foreign Information Manipulation and Interference (FIMI) in Taiwan’s 2024 national elections. Medium. Retrieved: https://medium.com/doublethinklab/artificial-multiverse-foreign-information-manipulation-and-interference-in-taiwans-2024-national-f3e22ac95fe7
- Dotson, J. (2023) Chinese information operation against Taiwan: The “abandoned chest piece” and “America skepticism theory”. Global Taiwan Institute. Retrieved https://globaltaiwan.org/wp-content/uploads/2023/08/OR_ASTAW0807FINAL.pdf
