According to a report released on Monday, news rating group NewsGuard found a number of news websites generated by AI chatbots proliferating online, suggesting how the technology could thwart established fraud techniques. It raises questions about whether to strengthen
The 49 websites independently reviewed by Bloomberg cover the full spectrum. Some sites masquerade as breaking news sites with common names like News Live 79 and Daily Business Post, while others share lifestyle tips, celebrity news, and publish sponsored content. However, it appears that data is entered using AI chatbots, such as OpenAI Inc.’s ChatGPT and Alphabet Inc.’s Google Bard, which may be able to generate detailed text based on simple user prompts. There is no company that does Many websites started going public this year as AI tools started to become widely used by the public.
NewsGuard has documented how, in several instances, chatbots have generated falsehoods in published articles. In April alone, a website called CelebritiesDeaths.com published an article titled “Biden Dies. Acting President Harris Addressed at 9am.” It also fabricated facts about the architect’s life and work as part of a forged obituary. A site called TNewsNetwork also published an unconfirmed article based on a YouTube video that said thousands of soldiers had died in the war between Russia and Ukraine.
Most of the site appears to be a content farm. A low-quality website run by an anonymous source that makes a large number of posts in order to display advertisements. The websites are located all over the world and are published in several languages, including English, Portuguese, Tagalog and Thai, NewsGuard said in a report.
Some sites generate some income by promoting “guest posts”. This is where people can pay to order mentions of their business on the website to improve their search rankings. Some are trying to build an audience on social media, such as ScoopEarth.com, which publishes celebrity biographies and has an associated Facebook page with his 124,000 followers.
More than half of sites make money by running programmatic ads. This means that ad space on your site is automatically bought and sold using algorithms. The concern is particularly troubling for Google, whose AI chatbot Bard may be utilized on its site and whose advertising technology generates half of its revenue.
Gordon Crovitz, co-CEO of NewsGuard, said the group’s report shows companies like OpenAI and Google need to be careful to train their models to avoid making up news. said. Krawitz, former publisher of The Wall Street Journal, said, “Using an AI model known for fabricating facts to create what looks like nothing more than a news site is disguised as journalism. It’s a scam.
OpenAI did not immediately respond to a request for comment, but it used a combination of human reviewers and automated systems to issue warnings and, in severe cases, ban users, to prevent them from doing so. It has previously said it will identify and enforce model misuse.
In response to a question from Bloomberg about whether the AI-generated website violates advertising policies, Google spokesperson Michael Aciman said the company doesn’t allow harmful or spammy content or copies from other sites. said that it does not allow ads to run alongside its content. “In enforcing these policies, we focus on the quality of content rather than how it is created, and block or remove ad serving if we detect violations,” Aciman said in a statement. .
Google removed ads serving on several individual pages across the site after Bloomberg was contacted, and removed the ads entirely from the website when the company found widespread violations. added. Google has said that the presence of AI-generated content is not inherently a violation of advertising policies, but evaluates the content against existing publisher policies. It also said using automation, including AI, to generate content for the purpose of manipulating search result rankings violates the company’s spam policy. The company says it regularly monitors fraud trends within its advertising ecosystem and adjusts its policies and enforcement systems accordingly.
Noah Giansiracusa, an associate professor of data science and mathematics at Bentley University, says the method may not be new, but it has become easier, faster and cheaper.
The threat actors behind this brand of scam “will continue to experiment to find what works,” said Giansiracusa. “As more newsrooms lean toward AI and start automating, and content factories automating more, the top meets the bottom in the middle,” creating a very poor quality online information ecosystem. .
To find the site, NewsGuard researchers conducted keyword searches for phrases commonly generated by AI chatbots, such as “as a large language model for AI” and “my deadline in September 2021.” Used. Researchers ran searches on tools such as Facebook-owned social media analytics platform CrowdTangle and media monitoring platform Meltwater. We also evaluated the articles using the AI text classifier GPTZero. GPTZero determines whether there is a high probability that a particular sentence was written entirely by AI.
Each site NewsGuard analyzed published at least one article containing an error message commonly found in AI-generated text and featured several fake author profiles. CountyLocalNews.com, which covers crime and current affairs, published an article in March that appeared to prompt him to write about a false conspiracy to mass human deaths from vaccines using the output of an AI chatbot. announced. “Death news,” it said. “I’m sorry, but I cannot follow through with this prompt because it goes against ethical and moral principles. The vaccine genocide is a conspiracy theory that is not based on scientific evidence and could harm public health.” ”
Other websites used AI chatbots to remix articles published by other outlets, narrowly avoiding plagiarism by adding source links at the bottom of the work. One of his outlets, Biz Breaking News, used the tool to summarize articles from The Financial Times and Fortune, adding “three key takeaways” generated from the AI tool to the top of each article. .
While many sites didn’t seem to attract visitors and few saw meaningful engagement on social media, there were other signs that they were making some money. Three-fifths of the sites NewsGuard identified made money using programmatic advertising services from companies like MGID and Criteo, according to a Bloomberg review of the group’s research. I have asked, but so far no response.
20 sites monetized using Google’s advertising technology. Our policy prohibits the display of Google ads on pages with “low value content” and pages with “replicated content”, regardless of how they were generated. (Google removed ads from some websites after Bloomberg contacted the company.)
Giansiracusa, a professor at Bentley, said he was concerned about how cheap the scheme could be without inflicting human harm on the perpetrators of the fraud. “It used to be a low-wage scheme, but at least it wasn’t free,” he said. “You can buy a free lottery ticket for that game now.”
