ChatGPT-like chatbots create fake news websites, spread misinformation

AI chatbots such as the widely popular ChatGPT have been used to create a slew of news and information websites. A considerable number of these websites have been published this year, thanks to the popularity of AI tools among the public. In its recently published report, news-rating group NewsGuard raised questions about how criminals may use this technology to enhance their existing fraud techniques.

ChatGPT-like AI bots were used to create 49 news websites, which were reviewed by Bloomberg. These websites pose as news outlets with generic names like Daily Business Post and News Live 79. Other websites publish sponsored content or offer lifestyle tips and celebrity news.

Interestingly, none of these websites acknowledge that their content is generated with the help of AI chatbots like OpenAI's ChatGPT or probably Alphabet's Google Bard. These chatbots are capable of generating text based on simple user prompts.

How AI chatbots are used to generate fallacies for published articles

NewsGuard found the AI chatbots created false information for articles that were published on these websites on multiple occasions. For instance, CelebritiesDeaths.com shared an article claiming US President Joe Biden had "passed away peacefully in his sleep" in April. A Yahoo News report stated that the fake article also indicated Biden had been succeeded by Vice President Kamala Harris.

Likewise, another website generated an obituary for an architect with fake details about their work and life. Also, TNewsNetwork constructed an unverified story about thousands of soldiers passing in the Russia-Ukraine war. This report was based on a YouTube video.

Most of these low-quality websites seem to be content farms that have been developed by anonymous sources. As expected, they aim to create posts that would generate advertising revenue. The NewsGuard report says these websites are located in different parts of the world. Also, they share content in multiple languages including Thai, Tagalog, Portuguese, and English.

How do these sites make money?

Some of these sites manage to generate income by advertising "guest posting." ScoopEarth.com and some other sites spared no effort to build a social media following. ScoopEarth's Facebook page has 124,000 followers. Over 50 per cent of these websites generate income by running programmatic ads, which are automatically bought and sold using algorithms.

This method of generating income poses a challenge for Google, whose ad technology generates revenue for half of the sites. Ironically, the search giant's AI bot Bard may have been used to create articles for these sites. So, NewsGuard co-Chief Executive Officer Gordon Crovitz says OpenAI, Google, and other companies behind AI chatbots should be cautious when it comes to training their models.

According to Crovitz, who is a former publisher of the Wall Street Journal, these companies should prevent their models from fabricating news. Crovitz further claims "Using AI models known for making up facts to produce what only look like news websites is fraud masquerading as journalism." OpenAI has previously clarified that it uses human reviewers and automated systems to prevent the misuse of its bot.

Still, cybercriminals have been coming up with new methods to exploit AI tools. For example, AI voice cloning tech was recently used in a kidnapping scam in the US. Similarly, scammers are using AI chatbots to create credible-looking phishing emails. Google spokesperson Michael Aciman told Bloomberg that the company restricts ads from running on spammy content or content that has been plagiarised from other websites.

Furthermore, Aciman explained that Google focuses more on the quality of the content rather than the creation process of the content when it comes to enforcing its policies. Also, the company immediately removes or blocks ads if they find any violations. The American tech giant recently removed ads from some sites where it detected pervasive violations.