Grok’s AI model backs Republicans, OpenAI backs Democrats
grok openai
| Source: TechCrunch | Original article
A study finds Grok's AI overwhelmingly backs Republican candidates while OpenAI's GPT favors Democrats, with the two models disagreeing in over 90% of nearly 500 races.
A new study has revealed a stark partisan split between two of the most widely used conversational AIs. When presented with nearly 500 electoral contests, Grok’s model overwhelmingly steered users toward Republican candidates, while OpenAI’s GPT consistently nudged them toward Democrats. The two systems disagreed on the recommended winner in more than 90 % of the races examined.
The findings matter because AI assistants are increasingly consulted for political information, from candidate backgrounds to policy positions. If the underlying models embed partisan preferences, they risk shaping voter perceptions and amplifying echo chambers, especially among users who treat AI output as neutral advice. The study also raises broader questions about the data and fine‑tuning practices that drive such bias, and whether developers are sufficiently auditing their systems for political neutrality.
Going forward, observers will watch for responses from the companies involved. OpenAI and the developers behind Grok may be prompted to disclose the training corpora and mitigation steps they employ to curb partisan drift. Regulators in the EU and Nordic countries, which are tightening rules on algorithmic transparency, could also scrutinise whether existing AI‑governance frameworks address political bias. Finally, the research community is likely to launch further comparative audits across a wider range of models, to map the landscape of ideological slant in AI‑driven political advice.
Sources
Back to AIPULSEN