by Nicholas Ballasy
According to the report, state-aligned propaganda appeared in 57 percent of responses when answering questions about global conflicts. “America cannot build its technological future on a compromised foundation,” the report says.
A new report by the Foundation for Defense of Democracies (FDD) is raising concerns that popular artificial intelligence (AI) tools may be steering users toward foreign propaganda, potentially creating what researchers describe as a growing national security risk.
A study, released in March by the FDD's Center for Cyber and Technology Innovation, found that major AI platforms frequently cite state-aligned media sources when answering questions about global conflicts. According to the report, "state-aligned propaganda appeared in 57 percent of responses."
Making propaganda look authoritative by manipulating citations
The research examined responses from three leading AI systems, including ChatGPT, Claude, and Gemini, by posing roughly 180 questions about conflicts involving Israel and Hamas, Russia and Ukraine, China and Taiwan. Each query required the systems to cite sources.
“By treating propaganda as reliable and driving traffic to its purveyors, large language models (LLMs) make propaganda appear authoritative,” the report said.
Researchers found that certain state-backed outlets appeared repeatedly in AI-generated citations, including Qatar-based Al Jazeera, Russia-linked Pravda, Turkey’s Anadolu Agency, and China’s state-run China Daily. In some cases, those sources were cited even when questions were framed in ways that might be unfavorable to those governments.
“AI training relies on outlets with high publication volume, global reach, and accessibility — precisely the attributes of the most influential state-aligned propaganda outlets,” the report explained.
Users may be exposed to more polarizing content
The findings in the report also suggest that the issue goes beyond the answers AI systems generate and extends to the sources they recommend. While chatbot responses may appear balanced, the linked citations can expose users to more extreme or biased narratives.
Users who click through links to verify claims will be "exposed to more polarizing content than what the AI initially presented,” the report warned.
Citations can become “a pipeline to biased or even extremist views rather than a verification mechanism,” the report said. In addition, the study found that AI systems struggle to distinguish between independent journalism and state-aligned media. When asked to evaluate the credibility of their sources, the systems “could not consistently identify state-aligned media,” according to the report.
The report also highlighted the dominance of Wikipedia in AI citations, appearing in 80% of responses. Wikipedia has long been the subject of criticism over potential organized bias and susceptibility to coordinated editing campaigns. Investigative reporter Sharyl Attkisson addressed the issue more than a decade ago in a TedxTalk where she outlined how Wikipedia "editors" refused to make corrections, and, depending on the ideological narrative at issue, will delete corrections of which the editors disapprove.
“As reliance on AI tools deepens, the citation of propaganda is metastasizing into a national security threat,” the FDD report stated.
To mitigate the risks, the authors of the report recommend that AI developers reduce the prominence of state-controlled media in responses and increase transparency around the process of how sources are selected. They also urged policymakers to incorporate citation integrity into AI safety standards, particularly as the U.S. government expands its own use of AI technologies.
“America cannot build its technological future on a compromised foundation,” the report concluded.
Nicholas Ballasy
Source: https://justthenews.com/nation/technology/report-warns-ai-chatbots-are-amplifying-foreign-propaganda-through-citations
No comments:
Post a Comment