Dem
53
Ties
2
GOP
45
Data centers aren't the only AI-related problem on the scene. So are AI chatbots. Many of them are highly partisan, and are delivering up lies and calling it unbiased news. A number of them are promoting very partisan websites, sometimes referred to as "pink slime," posing as neutral news sites.
This observation comes from an op-ed in Politico written by two people who work at NewsGuard, an organization that studies the reliability of news sites. What they did was ask seven leading AI chatbots questions about candidates and issues. In half the cases, the bots cited pink slime sites as reliable news sources. In a few cases, they were the only sources cited. The percentage of chatbot's responses that cited pink slime varied considerably. OpenAI's ChatGPT cited pink slime 71% of the time, Microsoft's Copilot 54%, Perplexity also 54%, Anthropic's Claude 50%, Google's Gemini 42%, Meta AI 38%, and xAI's Grok 29%. When asked about this, a Microsoft spokesperson said: "Copilot has been designed to ground answers in high-quality search results that balance safety, authority, and access to information." Did you expect: "Our bot is pretty stupid so it serves up fake news half the time"? Interestingly, left-leaning sites were cited 3x more than right-leaning sites. This may be due to the fact that there are many more postings to the former than the latter. Flooding the zone seems to work.
The prompts were designed to elicit responses about races in six battleground states: Alaska, Georgia, Iowa, New Hampshire, North Carolina, and Ohio. The sites cited as sources all had plausible-sounding names, like the PeachStatePost.com (R) and GranitePostNews.com (D). Parties have tried to shape the news since approximately when news was invented, but using highly partisan sources and presenting them as neutral reporters as news is new in America, although Russian disinformation campaigns go back a long ways.
The rise of partisan AI is important because increasingly many people use AI to get news and tend to believe what the bot says. Even people who don't want to use AI end up using AI. Many ordinary searches on Google now give results concocted by Google's Gemini bot rather than links to (reliable?) sources. NewsGuard identified 1,179 pink slime new sites, which is more than the 937 remaining daily newspapers in the U.S. If people get their information mostly from social media and bots regurgitating stories from pink slime sites run by political parties and their affiliates, this is not going to lead to a well-informed public. Pink slime and social media interact because many fake news sites buy ads on social media to increase readership. When voters see ads for fake news sites with plausible names on social media and then a chatbot cites them as a source, the voters see them as familiar and tend to swallow the pink slime whole, without questioning it.
Another observation NewsGuard made is that the fake news stories fall into two groups: positive news about the favored party and candidates and negative news about the disfavored ones. The bots tended to latch onto the negative stories more than the positive ones. Maybe the bots see the positive stories as unreliable "puff pieces" but the negative ones as good investigative journalism, which they certainly were not. Welcome to the future of news. (V)