Fact-check: “Studies from Stanford, MIT, and others, have found biased results and partisan leanings (bias) or sentiments in generative AI searches and AI Chat tools”
Verdict: Mostly True (85% confidence)
Multiple documented Stanford and MIT research outputs — including a Stanford GSB working paper (May 2025) finding nearly all LLMs perceived as left-leaning across 180,126 user assessments, a Stanford HAI/arXiv study (2023) confirming left-leaning tendencies in feedback-tuned models, and an MIT-linked EMNLP 2024 paper finding that optimizing reward models for truthfulness tends to produce left-leaning political bias — directly support the claim that studies from these institutions have found par…
This fact-check was conducted by SpinkillerAI, a non-partisan AI-powered accountability platform.
Related fact-checks
- Donald Trump Jr accepted a gift from a Russian oligearch — True
- European Union plans 200 million Euro investment in Greenland as a response to Trump's posturing — Mostly True
- Approximately 45% of the current Canadian population is made up of immigrants and the children of immigrants — Mostly True
- Recent job numbers increased faster than had been expected and are ahead of most analysts expectations — Partly True
- Medicare for all would cost trillions a year that Americans would pay through higher income taxes — Partly True