Reporting an Attribution Claim from Anonymous Sources


As the novel coronavirus spread through the United States in mid-March, so did mis- and disinformation about impending governmental action. On March 16, 2020, media outlets such as the New York Times and Vox published articles on rumors circulating via text message that warned of an imminent national quarantine or lockdown. According to the articles, there were several versions of the texts, beginning first on March 13 with messages claiming President Trump would invoke the Stafford Act, implementing a mandatory quarantine. As the hoax messages spread, the White House’s National Security Council tweeted on March 15 that the text messages were fake. The articles noted it was unclear who was behind the mass messages. 

On April 22, the New York Times published a follow-up story stating that American intelligence agencies had found “Chinese operatives helped push the messages across platforms.” In addition to text messages, the same rumors had been disseminated in encrypted messaging apps, and on Facebook, 4chan and various other platforms — using false personas to amplify the rumors. Two of the officials told the NYT they believed the Chinese operatives had not created but had spread existing rumors. The article reiterated that “the origin of the messages remains murky.”

Evidence of Attribution

The US officials who linked Chinese agents to the hoax text messages and social media posts chose not to share their evidence of attribution with the NYT, for concerns over exposing their sources and methods of monitoring Beijing’s covert activity. As a result, the article leaves open several questions on the details of the specific operation. The speculative claims that go unanswered include: 

  • Whether PRC-operatives created the lockdown messages or amplified existing ones.
  • Whether spies in the PRC’s diplomatic missions spread the fake lockdown messages.
  • Whether other “rival powers” helped spread the fake lockdown messages.

Each of these claims is missing concrete evidence or more specific details to understand why the operatives, spies and rival powers are believed to be linked to this operation. For example, which “rival powers” helped spread the fake lockdown messages, and were they working with the Chinese operatives?  

The article cites independent bipartisan research from a think tank and a national security advocacy organization as support for the officials’ broader concerns about China’s disinformation efforts. While this research looks at the PRC’s covert capabilities on messaging apps, it does not corroborate the claims that China was involved in this specific operation. For example, these reports do not mention the PRC’s capabilities with spreading disinformation via text messages. 

Why this Story is Challenging to Report

When disinformation about the federal government is spread to Americans amid a global pandemic, during a Presidential election year, using new methods difficult to track— it’s a major story. There is a public interest in identifying the individuals or organizations behind these messages, and the possibility that foreign actors might be involved only magnifies the importance of the story. 

Reporters are caught in a tough situation, however, when the entities making the attribution condition their participation on anonymity, and provide conclusions but not the underlying data or method. The findings will not be reproducible or confirmable by outside experts in such cases. Under what circumstances should a journalist grant this request for anonymity? How do you figure out whether you can trust your sources’ information, and if you do, how can you convince readers to trust the indirect attribution? 

In this case, the sources seeking anonymity were government officials, which added another layer of difficulty to the reporting. They claimed to have evidence of foreign agents circulating disinformation by text, but did not want to publicize the identification tactics for security reasons. 

Additionally, the false claims of a national lockdown spread not just on public platforms but also through text messaging and encrypted messaging apps — places where it is difficult to track disinformation. This means that platforms and researchers can at best provide journalists with insight into the provenance and spread of only a portion of the messages.

How the Press Covered Attribution

The NYT’s April 22 story carried an explosive claim — American intelligence agencies had assessed that Chinese agents helped spread the national quarantine rumors. The NYT, however, did not name the US officials, the agencies they worked in or the officials’ level of involvement with the investigations, noting only that the sources came from six different agencies, and that “some have spent many years analyzing China”. The article did not include specific evidence of Chinese participation in the influence operation.

The piece acknowledges the ongoing information warfare between the PRC and the US and that “American officials in the past have selectively passed intelligence to reporters to shape the domestic political landscape”, noting that some worry the current administration may leak intelligence findings to bolster an anti-China narrative. To address this concern for this particular story, the reporters used publicly available evidence to support the broader concerns expressed by their anonymous sources, pointing to a Chinese Foreign Ministry spokesperson and state-run media’s promotion of anti-US coronavirus conspiracy theories. 

To provide further context, the article situates this particular operation within the PRC’s larger disinformation campaign against the United States, noting its similarities to Russian-style tactics that aim to exacerbate existing political tensions and undermine the US public’s confidence in the government. 

The officials’ assertions of Chinese involvement were subsequently re-reported by multiple news outlets around the world. 

Assessing Press Coverage

Without names, sources, data and methods, readers may find it tricky to judge the credibility of the officials’ conclusions. Here are some tips for how communication of the attribution claim could have been further strengthened: 

  • Specify the confidence levels. The NYT spoke to multiple officials to ensure that the assessment was shared across multiple intelligence agencies. But how confident were these officials about the Chinese operatives’ involvement? Readers would benefit from knowing whether they had low, medium or high confidence in their conclusions. (See the Glossary for more on confidence levels.) The responsibility here doesn’t fall solely on journalists — government entities assigning attribution should offer a confidence level with their conclusions, which journalists should report to readers. But if the entity does not offer this information, the journalist should ask. 
  • Explain how the agencies and officials reached their conclusions. Even if the officials and agencies are not named, other general details can help readers assess the credibility of the findings. Did the officials directly take part in the threat monitoring and attribution process? Did the six agencies reach the same assessment independently of each other? Were they using the same set of evidence?
  • Avoid speculation without evidence. The April 22 article claims that “other rival powers might have been involved in the dissemination, too,” but could have strengthened this part by providing, or linking to, evidence for this claim. 
  • Consider that an actor engaging broadly in propaganda is not evidence of that actor engaging in a specific operation. China’s covert capabilities and previously attributed operations are discussed broadly throughout the article, but without clear attribution in this particular operation, it is difficult to link these capabilities to this singular campaign.