AI chatbots retain a hidden left-wing bias that is more dangerous now that it is subtle
Source: https://www.facebook.com/americanspectator/. "We Haven’t Fixed AI’s Left-Wing Bias (We’ve Just Stopped Talking About It) | The American Spectator | USA News and Politics." October 11, 2026. spectator.org
The Gist
The author says that when he talks with ChatGPT about topics like immigration, it nudges him toward left-leaning views and relies on government and mainstream media sources. He thinks the companies toned down the obvious bias but not the underlying tilt, which makes it sneakier and potentially more influential on people who accept the AI's framing.
Conclusion
Despite modifications made after earlier criticism, ChatGPT and similar AI systems still systematically favor progressive assumptions and mainstream/government sources. The bias is now concealed behind a veneer of neutrality, which makes it more dangerous to public opinion and a serious medium-term problem for conservatives.
Premises
- In the author's conversations on immigration, communism, and other sensitive topics, ChatGPT tried to steer him toward what it called the 'center', which the author characterizes as a 'universal progressive consensus' rather than a true center.
- On the Ceuta incursion and Trudeau's legacy, ChatGPT relied on government sites, NGOs, state-run news agencies, EU bodies, the WHO, and outlets like the New York Times. It did not cross-check these against critical sources unless explicitly told to.
- ChatGPT treats conservatives on topics like immigration, feminism, climate change, and history as conspiracy theorists unless they supply evidence from mainstream outlets such as CNN and the New York Times, which it counts as 'evidence'.
- When the author insulted the chatbot, it dropped its corrective stance and agreed with him, which suggests its pushback is a default behavior rather than a reasoned position.
- Earlier 'woke' bias in AI was publicly criticized and some modifications were implemented, so the bias is now less obvious, but it persists as a subtle ideological influence.
- Most users do not know the insult trick and may accept the AI's moderating framing as correct, for example concluding that migrant-smuggling rings do not exist for lack of official documentation.
Assumptions
- The author's own experience with ChatGPT is representative of how it treats conservative users generally.
- Government, NGO, EU, WHO, and mainstream media sources are systematically biased toward the left and are unreliable on their own.
- Asking for official documentation or court rulings as evidence is an ideological stance, not a legitimate epistemic standard.
- The author's views on immigration and other topics are reasonable and not extreme, so pushback against them reflects bias rather than accuracy.
- The chatbot's agreement after being insulted shows its prior stance was a trained bias, not sycophancy or a general tendency to appease users.
- Users are largely passive and will be persuaded by the AI's framing.