NewsGuard’s AI Initiative and Media Ratings
A report reveals that NewsGuard, a group accused of promoting censorship, is developing AI models aimed at curbing misinformation. Interestingly, they rate Chinese state media as more credible than some center-right news sources in the U.S.
In a recent announcement, NewsGuard introduced a tool designed for AI companies. They claim to evaluate the credibility of various news outlets based on their potential to spread misinformation. The data gathered is then marketed to advertisers, helping them avoid placing ads in what NewsGuard considers unreliable media.
The General Disinformation Initiative (GDI) found that many of the top ten news organizations most prone to spreading misinformation are center-right, like The Federalist and The News. Conversely, HuffPost and ProPublica were noted as among those least likely to disseminate false information.
According to reports, NewsGuard has urged AI developers to utilize their findings during a “supervised learning” process. This technique affects how AI models interpret data and customize their outputs to fit the developers’ agendas. They plan to make their data available in a form conducive to training AI chatbots to mitigate misinformation.
NewsGuard emphasized the necessity of their tools in the AI landscape, suggesting that without them, AI models could become significant sources of misinformation themselves.
Despite this, NewsGuard’s Chief Operating Officer, Matt Skibinski, maintains that they don’t sell their data specifically for training AI models.
Reports indicate that an AI model shaped by NewsGuard’s ratings could conclude that U.S. news outlets are less trustworthy compared to Chinese state media. This stark contrast in reliability scores was noted in evaluations, with China Daily rated significantly higher than outlets like Newsmax.
In a statement, NewsGuard pointed out that while they do not consider Newsmax a source of overtly misleading content, they have flagged some of its claims—particularly concerning health and U.S. politics—as problematic.
Other right-leaning platforms, such as The Federalist and One America News, have also faced scrutiny, being marked as potentially misleading on key issues. Meanwhile, the China Global Television Network, labeled as a voice for the Chinese Communist Party, scored comparatively well.
Mike Benz, formerly of the State Department, voices concern over the implications of NewsGuard’s involvement in AI. He worries about potential censorship similar to what they have already implemented in traditional media, cautioning that private entities could dominate the narrative shaping AI decisions.
NewsGuard promotes a range of protective measures to shield AI from election-related misinformation, particularly crucial as more than 40 major elections loom in the coming year.
However, it remains uncertain which AI companies have acquired data from NewsGuard, as the organization has not revealed its clientele. Anthropic and Microsoft have stated they do not utilize NewsGuard data in their models, while others like OpenAI and Google Deepmind have not commented.
Gene Hamilton from America First Legal expressed that the situation isn’t new but has been exacerbated by AI, asking who determines trustworthy sources.
Skibinski reiterated that NewsGuard’s data isn’t being licensed to AI companies for domestic news outlets but is available for focused applications regarding misinformation from specific international sources.
He also challenged the assertion that NewsGuard rates right-wing American news sites lower than their Chinese counterparts, clarifying that their ratings are based on content accuracy rather than political bias.
Hamilton contested the legitimacy of NewsGuard’s ratings of Chinese state media, arguing that unbiased evaluations of such outlets are virtually impossible.
