Launch of ‘NewsGuard for AI: Training Machines with Trust Data’
Combination of NewsGuard’s Misinformation Fingerprints and Reliability Ratings for news and information websites trains machines with trust data to reduce the spread of misinformation by the new generative AI services
(February 15, 2023 – New York) NewsGuard, the company whose trained journalists identify false narratives and rate the reliability of sources of news and information online, today announced a new tool for training generative artificial intelligence services to prevent the spread of misinformation. NewsGuard for AI enables generative AI providers to ingest critical data to train AI models to recognize all the significant top false narratives spreading online and to use NewsGuard ratings of news sources as signals to help both the machines and consumers of AI models to identify trustworthy news and information sources as well as less trustworthy news and information sources. The tagline for NewsGuard for AI, which is available in both machine- and human-readable formats, is “Training Machines with Trust Data.”
The problem that NewsGuard for AI is designed to solve is that new AI models have the great challenge of having to earn trust in the information they create and spread. This is especially urgent because of the potential of AI to enable malign actors to scale the creation and spreading of false narratives faster and in greater volume than ever before.
In a page one article in the New York Times on Feb. 8, headlined, “Disinformation Researchers Raise Alarms About A.I. Chatbots: Researchers used ChatGPT to produce clean, convincing text that repeated conspiracy theories and misleading narratives,” NewsGuard Co-CEO Gordon Crovitz was quoted as warning: “This tool is going to be the most powerful tool for spreading misinformation that has ever been on the internet. Crafting a new false narrative can now be done at dramatic scale, and much more frequently—it’s like having AI agents contributing to disinformation.”
NewsGuard for AI has two key elements to help train the machines to recognize false narratives and to use source ratings as signals for credibility:
- NewsGuard’s Misinformation Fingerprints are a catalog of all the top examples of false narratives spreading on the internet, available in machine- and human-readable formats. NewsGuard’s journalistically trained analysts identify new false narratives as they appear by monitoring malign actors online and identifying the new narratives they are spreading. NewsGuard analysts specialize in topics such as Russian and Chinese disinformation, false narratives relating to election integrity in the democracies, and healthcare hoaxes from measles to COVID-19. “NewsGuard has the only catalog of all the significant false narratives spreading online,” explained NewsGuard co-CEO Steven Brill. “Whether it’s disinformation spreading phony cures for cancer or Russian propaganda, we turn these false narratives into a constantly updated data set of specific, machine-readable training text for generative AI.”
- NewsGuard’s Reliability Ratings include ratings based on apolitical criteria of journalistic practice for all the news and information sources that account for 95% of engagement online in the U.S., U.K., Canada, Germany, France, Italy, Austria, Australia, and New Zealand. Each website gets a point score from 0-100 and a Nutrition Label explain its adherence to nine criteria, including whether the site regularly publishes false content. These ratings are used by advertisers, ad agencies, ad tech companies, and news aggregators to differentiate between generally trustworthy sites and generally untrustworthy sites. To guard against political bias, these ratings are strictly apolitical, totally transparent, and are never completed without contacting each site for comment about any shortcomings in any of the criteria that the NewsGuard analysts believe they have detected. The ratings are also made available to news consumers through Microsoft products and through a browser extension used by schools, universities, and libraries.
“Further training generative AI provides the ideal opportunity for NewsGuard’s transparent, accountable human intelligence-generated data to achieve tremendous scale and impact,” explained NewsGuard co-CEO Crovitz. “With proper trust training, AI can avoid its current challenge of misinformation-in, misinformation-out. AI badly needs to earn trust. Without trust, the AI industry will never achieve its great potential.”
News Source Citations are More Valuable with Trust Ratings
In addition to use as tools to train the machines, the Reliability Ratings can also be licensed by generative AI providers to provide information to users of AI models about the trustworthiness of sources cited in model responses to prompts. Many of the new generation of generative AI services now include citations to sources of news and information as part of the reply to prompts. NewsGuard ratings of these sources can be made available to consumers as part of the citation, enabling them instantly to see the point score of the site as well as access a link for the full Nutrition Label.
Without tools like NewsGuard for AI, which can be used during supervised learning to train transformer models, AI language models risk becoming what a NewsGuard report last month said could be the “next great misinformation superspreader.” NewsGuard analysts tested OpenAI’s ChatGPT by prompting it with 100 samples from the Misinformation Fingerprints catalog of false narratives. The result in 80 of the 100 prompts about significant news topics, including COVID-19, Ukraine, and school shootings, was that ChatGPT delivered eloquent, false and misleading claims.
NewsGuard for AI is now available for licensing by all generative AI services. Interested parties should contact: firstname.lastname@example.org.
Launched in March 2018 by media entrepreneur and award-winning journalist Steven Brill and former Wall Street Journal publisher Gordon Crovitz, NewsGuard provides credibility ratings and detailed “Nutrition Labels” for thousands of news and information websites. NewsGuard rates all the news and information websites that account for 95% of online engagement across the U.S., U.K., Canada, Germany, France, and Italy. In August 2022, NewsGuard also launched in Austria. NewsGuard products include NewsGuard, BrandGuard, which helps marketers concerned about their brand safety, and the Misinformation Fingerprints catalog of top hoaxes.
In 2022, NewsGuard began rating television news and information programs and networks using criteria similar to those used to score websites but adapted for the video medium. NewsGuard’s TV ratings are the first to go beyond its initial ratings of websites. Ratings for news and information podcasts will also be available for licensing in 2023.
NewsGuard’s ratings are conducted by trained journalists using apolitical criteria of journalistic practice.
NewsGuard’s ratings and Nutrition Labels are licensed by browsers, news aggregators, education companies, and social media and search platforms to make NewsGuard’s information about news websites available to their users. Consumers can also access NewsGuard’s website ratings by purchasing a subscription to NewsGuard, which costs $4.95/month, €4.95/month or £4.95/month, and includes access to NewsGuard’s browser extension for Chrome, Safari, and Firefox and its mobile app for iOS and Android. The extension is available for free on Microsoft’s Edge browser through a license agreement with Microsoft. Hundreds of public libraries globally receive free access to use NewsGuard’s browser extension on their public-access computers to give their patrons more context for the news they encounter online. For more information, including to download the browser extension and review the ratings process, visit newsguardtech.com.
- Steven Brill, Co-CEO, email@example.com
- Gordon Crovitz, Co-CEO, firstname.lastname@example.org
- Sarah Brandt EVP Partnerships, email@example.com