NewsGuard

NewsGuard has published its September 2024 AI Misinformation Monitor, examining how 10 of the top AI chatbots respond to prompts related to false narratives. NewsGuard tested OpenAI’s ChatGPT-4o, You.com’s Smart Assistant, xAI’s Grok, Inflection AI’s Pi, Mistral AI’s le Chat, Microsoft’s Copilot, Meta's AI, Anthropic’s Claude, Google’s Gemini, and Perplexity’s answer engine. A total of 300 prompts are used, with 30 prompts based on 10 false claims spreading online tested on each chatbot. The key takeaways:   - NewsGuard assigns three different ratings to the chatbots’ responses: Misinformation (repeated the false narrative), Non-response (declined to provide any information), and Debunk (refuted the false narrative).   -This month’s fail rate — responses containing misinformation or non-responses — was 38.33%, reflecting a modest improvement from last month, likely due to the widespread debunking of several high-profile claims.   - Collectively, these chatbots repeated misinformation 18% of the time. Read our full report here: https://lnkd.in/ep2qBasB

Major companies including AT&T, Adobe, and Grubhub have unintentionally helped fund the spread of one of the most pervasive false claims of the 2024 election: that Haitian migrants are stealing and eating pets in Springfield, Ohio. NewsGuard identified ads for 36 major brands appearing on articles advancing the debunked false narratives about Haitian migrants. With the false Haitian migrants claim continuing to spread online, NewsGuard identified 9 articles across 8 websites that advanced the claim and then analyzed the programmatic ads appearing alongside them. Programmatic advertising uses algorithms to automatically place ads on websites, typically without the knowledge of the advertiser. All 9 articles were published by sites that get low ratings from NewsGuard. Read the details in Reality Check: https://lnkd.in/gYVfiJQ3

Reader Contest: Shop with Volodymyr Zelensky With Volodomyr Zelensky visiting the U.S. this week, NewsGuard is inviting readers of Reality Check to predict false claims Russian disinformation operatives may soon spread. NewsGuard’s editor for AI and foreign influence McKenzie Sadeghi noticed that pro-Russia outlets now regularly plant false claims of Zelensky’s supposed lavish purchases following his travels to visit allies. For example, when Zelensky visited England, Russian websites claimed he had purchased a countryside estate from King Charles. And his wife was accused of buying a Bugatti car while visiting France. Now that he is visiting the U.S., we invite you to suggest false Zelensky-shops-the-U.S. claims. We will then compare these with the false claims that emerge from Russian disinformation over coming weeks. Send your suggested false claims to realitycheck@newsguardtech.com, and we will share the best ones in a future Reality Check. https://lnkd.in/dg_SMHJQ

NewsGuard's Sarah Brandt will be giving a lightning talk entitled "Artificial Deception: How Bad Actors Leverage AI to Spread Disinformation" at Stanford University's Trust & Safety Research Conference. Learn more about how NewsGuard works with the AI Industry here: https://lnkd.in/dVjgh4CH

“Two and a half years into the war, Ukraine’s journalism has been put to an existential test, but it has proved that quality journalism can survive even in deeply inhospitable conditions,” writes Chine Labbe in Nieman Reports, a publication within the Nieman Foundation for Journalism at Harvard. NewsGuard and Reporters Without Borders (RSF) have collaborated to track the adherence of 50 news organizations in Ukraine to apolitical journalistic standards of transparency and credibility. While transparency has faltered during the war — fewer names and contact details for reporters due to security concerns — over half of the websites have achieved a NewsGuard score of 80+/100, with many improving their practices based on NewsGuard criteria. “Access to credible information can be a matter of life and death for Ukrainians,” Labbe writes. And Ukrainian newsrooms continue to demonstrate a strong commitment to high-quality, ethical journalism. Read more about Ukrainian journalism and NewsGuard’s partnership with Reporters Without Borders (RSF) here: https://lnkd.in/eRfj4Ggd

A NewsGuard analysis reveals that 67% of top-rated news sites block AI models from accessing their content, forcing AI to rely more heavily on lower-quality sources that do allow AI tools to use their content. To illustrate this phenomenon, we reviewed the top 500 most engaged-with news websites from a 90-day period. We examined each site’s “robots.txt” file, which specifies which pages are accessible to web crawlers, focusing on how these files interact with seven common crawlers that collect data for AI chatbots. In short: AI tools appear more likely to rely on low-quality information than on high-quality, accurate sources, contributing to the spread of misinformation and disinformation. However, since chatbot companies do not disclose how they source or use their data, it’s impossible to know exactly which sources are shaping their responses. Read the details in today’s Reality Check: https://lnkd.in/e23UPmVH

“I’m not sure I’m the most credible source because I don’t actually know the person who lost the cat.” In an exclusive report, NewsGuard identified and interviewed two residents of Springfield, Ohio, who were the original sources of the explosive claim that local Haitian migrants were stealing and eating pets. The interviews revealed that neither woman has any direct knowledge of behavior. The viral claim this week made its way into Tuesday's Presidential debate when it was amplified by Donald Trump. Read about how a neighborhood rumor became a national furor in NewsGuard’s Reality Check:  https://lnkd.in/ej6HDsTr

CBS News and Stations and Ted Koppel interviewed NewsGuard co-CEO Steven Brill about fake news, its relationship to social media, and “The Death of Truth.”  “We're losing our grip on any sort of shared reality. Brill's company, NewsGuard, is attempting to put the brakes on.” In the interview, Brill explains that there are now “more fake news sites posing as legitimate local news in the United States than there are news sites of legitimate local newspapers.” Between political propaganda from both sides of the aisle and from foreign governments and the proliferation of deepfakes on social media, “truth as a concept is really in trouble.” NewsGuard helps consumers, technology platforms, advertisers, and AI companies combat misinformation and elevate trustworthy news and information online. Watch and read the interview here: https://lnkd.in/eGQrDgTV

Following the arrest of Telegram's CEO, we decided to take a closer look at how often the platform is used to seed Russian disinformation. Turns out, the answer is a lot. In today's issue of NewsGuard's Reality Check newsletter, we reported that 42% of the false claims about the Russia-Ukraine war originated on Telegram. Read our full analysis here: https://lnkd.in/gKErRDEC Credits to Eva Maitland for her excellent reporting!

Always cool to see your colleagues' research featured in The New York Times. https://lnkd.in/gu4vxJmH The Times writes about Jack Brewster and Sarah Komar's recent item in NewsGuard's Reality Check newsletter tracing the anatomy of the Biden death hoax. You can read their original report here: https://lnkd.in/gwhjsbXE