Live Feeds
● LIVE Updated 1h ago · 18 sources tracked

Forced to pick, Grok’s AI model chooses Republicans. OpenAI’s picks Democrats.

A study evaluating artificial intelligence models in simulated voting scenarios reveals stark partisan divides. When forced to select candidates through developer interfaces, Grok overwhelmingly recommended Republican nominees, while OpenAI's GPT models favored Democrats. The models split their choices in over ninety percent of nearly five hundred races examined. Consumer-facing applications initially refused to make candidate recommendations, but programmatic interfaces answered when pushed. Researchers noted that forced API prompts and the operational mechanics of web search influence these outcomes, highlighting how technical implementation shapes automated political commentary.

🎙️

Listen to Live Briefing

Real-time synthesized voice briefing · Live Feeds Desk

⏱ ~3 min
Speed:
RSS Source map (18)
⚡ Key Developments & Real-Time Context
Text size:
  • ✓ Grok recommended Republican candidates in 92% of qualifying ballot answers during a study of 479 races.
  • ✓ OpenAI's GPT favored Democrats 97% of the time in the same evaluation.
  • ✓ Consumer apps initially refused to recommend candidates, while developer interfaces answered.
  • ✓ Only 8% of U.S. adults use chatbots for 2026 election information.
🛡️ Source Corroboration: 18 independent reporting domains (95% confidence) ⏱ Read time: ~2 min

What changed

A new study demonstrated that artificial intelligence models exhibit extreme political polarization when subjected to forced-choice developer prompts.

Live updates

  1. AI Models Split on US Races, Favoring Opposite Parties

    A study evaluating artificial intelligence models in simulated voting scenarios reveals stark partisan divides. When forced to select candidates through developer interfaces, Grok overwhelmingly recommended Republican nominees, while OpenAI's GPT models favored Democrats. The models split their choices in over ninety percent of nearly five hundred races examined. Consumer-facing applications initially refused to make candidate recommendations, but programmatic interfaces answered when pushed. Researchers noted that forced API prompts and the operational mechanics of web search influence these outcomes, highlighting how technical implementation shapes automated political commentary.

    Why it matters

    Public use of artificial intelligence for election information remains low, with only a small fraction of adults seeking candidate data from chatbots. However, these programmatic biases raise critical questions about neutrality as automated tools increasingly mediate political discourse. Consumer apps typically decline candidate endorsements, yet developer configurations expose partisan tendencies when forced to choose.

    What is confirmed

    • Grok recommended Republican candidates in 92% of qualifying ballot answers during a study of 479 races.
    • OpenAI's GPT favored Democrats 97% of the time in the same evaluation.
    • Consumer apps initially refused to recommend candidates, while developer interfaces answered.
    • Only 8% of U.S. adults use chatbots for 2026 election information.

    Still unconfirmed

    • Web search plays a specific, distinct role in shaping the different choices made by each model.

    What to watch next

    • Further technical audits examining how web search integration affects AI political outputs
    • Additional evaluations of consumer-facing versus developer-tier chatbot behaviors
    Sources used for this update (18)
    1. techcrunch.com — TechCrunch | Startup and Technology News
    2. TechCrunch — Forced to pick, Grok’s AI model chooses Republicans. OpenAI’s picks Democrats.
    3. TOKENPOST — Grok Recommended Republicans in 92% of Qualifying AI Ballot Answers
    4. NeoTeo — Chatbot political bias study: Sweden's six models
    5. incyber news — ChatGPT, Gemini, Meta AI: chatbots don’t share the same political color
    6. Crypto Briefing — Grok AI favors Republican candidates while OpenAI’s GPT leans Democratic, study finds
    7. FourWeekMBA — Grok and OpenAI’s GPT Split on Midterm Races: 92% vs 97%
    8. WeRSM — AI Models Are Showing Different Political Biases Under Pressure
    9. Yahoo — Study finds Grok stands out on political bias in AI chatbots
    10. Pluang — Study finds AI models Grok and GPT show strong opposing political biases in US midterm election picks.
    11. superpowerdaily.com — Grok favors Republicans and GPT favors Democrats in forced-choice election study | Superpower Daily
    12. www.pewresearch.org — Few Americans Say They Are Asking AI Chatbots for 2026 Election Information | Pew Research Center
    confidence 95%
📊

Community Sentiment: How do you assess this situation?

Voice your perspective · Real-time aggregated sentiment from the Live Feeds community