UCLA Center for European and Russian Studies — Even a poorly made deepfake that fools no one can undermine public trust in information, according to researchers Elizabeth Ashley Fox-Jensen and Sue-Ellen Njekwa Ryberg, who discussed the growing threat of AI-generated disinformation at a talk hosted by the UCLA Center for European and Russian Studies on October 13, 2026.
Their presentation, “Weaponized Uncertainty: Deepfakes, Democratic Resilience and the Scandinavian Test Case,” explored how synthetic media can erode democratic communication by making authentic information easier to dismiss as fabricated.
Fox-Jensen, visiting assistant professor of information studies at UCLA and principal investigator of the Deepfake Democracy Project, and Njekwa Ryberg, a researcher in communication and policy at Uppsala University, presented findings from an interdisciplinary research initiative conducted at Malmö University in Sweden.
The project brings together faculty and student researchers from computer science, international relations, communication, digital humanities and design to investigate the creation, detection and democratic implications of political deepfakes.
“A deepfake that convinces nobody, even a bad deepfake, is still doing something,” said Fox-Jensen. “Even a bad one still does its job.”
A deepfake is an AI-generated or manipulated image, video or audio recording that portrays a person saying or doing something they did not actually say or do. Although the technology has become increasingly sophisticated, the researchers argued that its greatest danger may lie not in convincing people to believe false information, but in making them question whether any information is authentic.
The researchers drew on the concept of the liar's dividend, introduced by legal scholars Robert Chesney and Danielle Citron in 2019, which describes how the existence of synthetic media allows public figures to dismiss genuine evidence as fabricated.
Building on this theory, the Deepfake Democracy Project developed the concept of ambient skepticism, which Fox-Jensen described as a broader condition in which people become increasingly doubtful of information circulating through digital media.
“Our term that we coined in our group is ambient skepticism, and that means just making everybody skeptical of everything,” she explained.
The distinction is significant because skepticism can persist even when individual deepfakes are recognized as false. In such an environment, the burden of proving authenticity increasingly falls on journalists, political communicators and other professionals responsible for maintaining public trust.
Sweden as a test case
Sweden provides a particularly revealing setting for the research because of its historically high levels of institutional trust, strong press protections and longstanding investment in psychological defense.
According to figures presented by the researchers, 77% of users trust Sweden's public service broadcasters, and the country ranked fifth globally for press freedom in 2026. Sweden also established its Psychological Defence Agency in January 2022 to strengthen societal resilience against disinformation and foreign influence.
Yet these protections have not prevented the harmful circulation of AI-generated content.
Fox-Jensen highlighted the January 2026 case involving Swedish Deputy Prime Minister Ebba Busch, whose likeness was used to generate sexualized images through Grok, the AI system integrated into Elon Musk's social media platform X.
The incident illustrated both the gendered nature of deepfake abuse and the difficulty of regulating rapidly developing technologies.
Research presented at the talk indicated that women account for the overwhelming majority of subjects depicted in sexually explicit deepfake imagery. Fox-Jensen argued that this form of targeting has implications beyond individual reputational harm, potentially discouraging women from participating in political and public life.
The researchers also traced the regulatory response to the Grok controversy, noting that restrictions on image generation followed public pressure and that European legal responses developed over subsequent months.
A Dutch civil court imposed a penalty of €100,000 per day for noncompliance with an order concerning the generation and distribution of nonconsensual intimate imagery. Meanwhile, European policymakers continued debating restrictions and enforcement mechanisms for AI-generated sexual content.
For Fox-Jensen, the case revealed a central contradiction: new technologies can produce and distribute harmful content almost instantaneously, while legal and institutional safeguards operate much more slowly.
What journalists and political communicators are experiencing
Njekwa Ryberg presented findings from 13 semi-structured interviews conducted in February and March 2026 with five Swedish journalists and eight political communicators.
Her research examined how professionals responsible for informing the public understand deepfakes, approach verification and assess threats to democratic trust.
“Checking one fake video is possible. Checking a constant background doubt about everything is not,” she said.
The interviews revealed different concerns among the two professional groups. Journalists tended to emphasize the risks to their credibility if manipulated material entered news coverage, while political communicators were more concerned with institutional preparedness and the potential use of deepfakes as instruments of information warfare.
“Threat actors move faster than societies can respond,” said one interview participant, a specialist in total defense and resilience, whose identity was anonymized for the research.
Another participant, a senior researcher and policy advisor, used the Swedish term *verklighetsapati*, or “reality apathy,” to describe a situation in which people become so uncertain about information that they gradually abandon efforts to distinguish truth from falsehood.
Njekwa Ryberg connected this observation to the broader concept of ambient skepticism.
Her interviews also identified the disproportionate targeting of women in positions of power. One journalist emphasized how manipulated images can be used to demean women, while a political advisor warned that a convincing deepfake broadcast by a trusted public media institution could cause immediate damage.
The project also included a small survey of 28 Malmö University students. Of those surveyed, 71% reported having encountered a deepfake online, while 57% felt only somewhat confident in their ability to identify one. Half had never heard of the Ebba Busch case.
Although the sample was limited, the findings suggested that exposure to synthetic media does not necessarily translate into awareness of its political implications or confidence in detecting manipulation.
Beyond media literacy and detection
The researchers emphasized that teaching individuals to recognize manipulated content, while valuable, cannot fully address a problem embedded in the broader information environment.
Fox-Jensen described the project's work on deepfake detection technologies, including an assessment of tools based on their accuracy, explainability, costs and ability to perform across different contexts.
Some systems that perform well in controlled testing environments, she noted, are less reliable when applied to unfamiliar material or different populations. Many also remain too technically complex for ordinary users.
The project's design and mitigation research therefore explored ways to make verification more accessible through interfaces that could provide warnings directly on smartphones or at the operating-system level.
“We propose that it's on an operating system level,” Fox-Jensen explained, comparing the approach to accessibility features already built into mobile devices.
Rather than requiring individuals to search for a separate detection service, such tools could alert users to potentially manipulated content as they encounter it.
The researchers cautioned, however, that technological detection alone would not eliminate uncertainty. Detection systems themselves can produce imperfect results, and increasingly sophisticated AI tools continue to challenge their effectiveness.
The discussion also addressed the European Union's evolving regulatory framework, including the AI Act, Digital Services Act and European Democracy Shield.
Fox-Jensen described competing approaches to AI governance, with European policymakers simultaneously treating the technology as a security risk and an important source of economic innovation.
The research identified four possible directions for democratic responses: greater transparency, stronger legal accountability for harmful content, safeguards built into technology platforms and increased investment in the institutions that establish and verify public information.
None, the researchers argued, would be sufficient on its own.
Read the researcher biographies and more about the presentation.