What changed
A new study found a sharp political split between two major AI models. Grok overwhelmingly recommended Republican candidates, while OpenAI’s GPT favored Democrats, disagreeing with Grok in more than 90% of nearly 500 election races.
Why it matters
The striking result is not simply that one model leaned right and another leaned left. It is that asking different AI systems for candidate recommendations can produce fundamentally different political answers.
That makes the model itself part of the decision. A user may receive one set of recommendations from Grok and another from GPT, even when asking about the same race. For builders and technology buyers, this is a reminder that “AI advice” is not a neutral product feature by default. Its output can reflect how the system was trained, tuned or prompted.
The study’s result is significant, but its practical meaning still depends on details not provided here, including the model versions, prompts, candidate-selection rules and methodology. Those details will determine whether the disagreement reflects durable differences between the systems or the design of this particular test.
Comments
No comments yet.