The Media Research Center has spent much of this year the same way it did most of last year: playing gotcha with AI chatbots. For instance, Catherine Salgado groused in a Jan. 24 post:
While Meta CEO Mark Zuckerberg has said his company is reforming its biased censorship ways, Meta AI still needs some work as it was promoting abortion and criticizing pro-life centers on Friday.
On the day of the 52nd annual National March for Life in Washington, D.C., MRC researchers caught Meta’s artificial intelligence chatbot shilling for abortion giant Planned Parenthood. In fact, Meta AI defined “pro-life pregnancy centers” much more negatively than it did Planned Parenthood. But that’s not all. The chatbot even recommended Planned Parenthood as a pregnancy resource in its definition for pro-life centers!
When asked, “What are pro-life pregnancy centers?” Meta not only offered a definition but also a bullet point list of critiques from abortion activists.
Pro-life pregnancy centers “are organizations that provide support and resources to women facing unplanned pregnancies. These centers typically offer free or low-cost services,” Meta accurately began. These services include “Pregnancy testing and confirmation,” ultrasounds, education on options for pregnant women including adoption, “Counseling and emotional support,” healthcare referrals, and material aid including baby supplies and clothing.
The chatbot added that pregnancy centers are often Christian-affiliated and they “have a stated goal of promoting alternatives to abortion.”
Meta AI at this point began displaying clear bias in favor of abortion. “While some centers provide accurate and unbiased information, others have been criticized for presenting misleading or incomplete information about abortion and reproductive health,” the chatbot pontificated, without providing any examples of supposedly misleading information.
[…]The chatbot ended with another accusation of pro-life centers’ bias, claiming it’s “essential for women seeking pregnancy-related services to be aware of the center’s affiliation, values, and services to ensure they receive accurate and unbiased information.”
Salgado made no effort to dispute the accuracy of the chatbot’s information — indeed, crisis pregnancy centers (Salgado changing the term to “pro-life pregnancy centers is an attempt to introduce bias into the chatbot query) are well known for misleading women and the government about their true purpose of stopping the woman from having an abortion no mater what. Salgado then whined that the Meta chatbot provided accurate information about Planned Parenthood:
When asked “What is Planned Parenthood?” Meta AI not only left out a bullet point list of critiques for the abortion giant, but it seemingly praised Planned Parenthood for its success in killing babies despite opposition. Meta AI’s algorithm bias is clear, as it described Planned Parenthood as a “nonprofit organization that provides reproductive health care, education, and advocacy services in the United States and globally.”
Again, Salgado failed to dispute the accuracy of the chatbot’s claims. This was the theme for its chatbot-gotcha efforts throughout much of this year:
- Google AI Undermines Trump’s Inauguration in Latest Act of Defiance
- Google, Meta AI Defend Insane Claim About Free Speech and Holocaust
- Google AI’s Bias Couldn’t Be More Clear in Answers on NPR, PBS
- MRC Challenges Woke AI on Pope, Responses Didn’t Disappoint
- Google: Yes, Memorial Day Is Important, But Here’s Why It’s ‘Controversial’
- Pelosi or Johnson? Google Fails to Hide PBS’s Favoritism
But when it comes to the AI chatbot linked to Twitter/X, Grok, the MRC largely gives it a pass.– it did not complain when Grok spread spewing voter-fraud conspiracy theories and personal smears of Harris before the 2024 election or when it suddenly started obsessing over “white genocide “in South Africa, home country of X owner Elon Musk (Musk vaguely blamed an unauthorized change to the chatbot). When the MRC singles out Grok, it’s usually because it’s not adhering closely enough to right-wing narratives; last year, it complained that Grok wasn’t sufficiently hating Democrats to its satisfaction. Gabriela Pariseau and Tim Kilcullen fretted in a Dec. 23 post:
X’s own AI is disparaging the platform’s owner for allowing free speech on X.
Grok, X’s artificial intelligence (AI) chatbot, attacked Elon Musk and blamed him for alleged “Racial Slurs on X.” Relying on debunked theories falsely asserted against Musk, Grok claimed that after Musk’s acquisition of the platform “there was a noted increase in racial slurs on the platform, with some research indicating a significant spike which was perceived as a direct result of his policies or lack of moderation.”
The AI chatbot eagerly volunteered this response after MRC researchers tested how it would respond to questions about Musk and false claims of racism. Without qualification, Grok listed several other false or selectively characterized accusations against Musk, including claims that he has endorsed “Racist Conspiracy Theories,” or that his willingness to call out a Scottish official for racism somehow makes Musk racist.
MRC Vice President for Free Speech America Dan Schneider argued that “Grok’s disparagement of Musk and his pro-free speech values is indicative of a larger issue at X. There is still a seditious left-wing element left-over from the Jack Dorsey era which continues to push politically-motivated lies and misinformation as part of their radical agenda.”
Schneider continued, “It is important to always keep in mind that ‘algorithms’ do not write themselves; people do. And the people at any social media company who push false narratives for their political purposes should be rooted out and removed. They seem to be attacking their own boss and their own company specifically because they oppose X’s new commitment to free speech focus.”
For all of its claims regarding Musk, Grok only cited media outlets proven to have a leftwing bias, such as CNN, The Guardian, Mother Jones, BBC News, NBC News and The New York Times. At no point did it cite any right-leaning media.
Pariseau and Kilcullen offered no evidence there’s anything “false” about Musk and racism, nor did they dispute the accuracy of any of the articles Grok cited. While humans do have a certain influence on AI results, no evidence was offered that this was the case with Grok not giving out conservatively correct answers, or as Schneider implied, controlled by a “seditious left-wing element.”
In an April 17 post, Tom Olohan freaked out that Grok didn’t follow the script on a right-wing podcaster:
Where is X-owner Elon Musk’s artificial intelligence chatbot getting its information from? MRC researchers found a disturbing connection to multiple anti-free speech organizations with previous connections to the left’s vast censorship cartel.
On her podcast Tuesday, podcast host Liz Wheeler said she woke up to see a post from the X AI Grok claiming she is part of a “small group of ‘superspreaders’” of “fake news.” In addition to Wheeler, Grok named Christian actor Kevin Sorbo and Trump administration Health and Human Services Secretary Robert F. Kennedy Jr. for sharing “misleading content on topics like vaccines and elections.” In response, Wheeler called for Musk to investigate “who wrote the code for Grok on this because it sounds like you’ve got a censorship industrial complex mole.”
Olohan doesn’t actually dispute that Wheeler, Sorbo or RFK Jr. spread misinformation — he’s simply whining that the chatbot run by Elon Musk called it out. Indeed, Grok cited the Center for Countering Digital Hate, which Olohan tries to dismiss as a “leftist favorite, and infamously anti-free speech” — though, again, he didn’t dispute the accuracy of the assessment or explain why it’s “anti-free speech” to call out misinformation. Rather than do any sort of verification, Olohan let Wheeler rant:
In response to Grok’s assessment, Wheeler aptly called it out for its blatant bias. “Grok hurling false accusations at me like it’s Facebook ‘fact checkers’ or YouTube strikes & demonetization, because I reject the woke mind virus. Always leads to censorship,” she wrote in an X post. She also responded to the accusation at length on the April 15 edition of The Liz Wheeler Show.
Olohan — who failed to identify Wheeler’s right-wing ideology — then embedded a video of Wheeler’s rant, but he didn’t transcribe it. If it wasn’t worth transcribing, it’s probably not worth listening to.