POLICY · REGULATION · #3012
Research finds mainstream chatbots often give incorrect or partisan midterm election info
WIRED summarizes research from Forum AI and NewsGuard showing that mainstream chatbots — including OpenAI’s ChatGPT, Anthropic’s Claude, Google’s Gemini and xAI’s Grok — frequently provide outdated, hallucinated, or partisan-sourced answers about the US midterms. Forum AI reported almost 40% of election-related answers contained at least one factual error, while NewsGuard found chatbots cited so‑called “pink slime” sites roughly half the time (ChatGPT >70%, Grok ~30%).
KEY POINTS
- WIRED summarizes research from Forum AI and NewsGuard showing that mainstream chatbots — including OpenAI’s ChatGPT, Anthropic’s Claude, Google’s Gemini and xAI’s Grok — frequently provide outdated, hallucinated, or partisan-sourced answers about the US midterms.
- Forum AI reported almost 40% of election-related answers contained at least one factual error, while NewsGuard found chatbots cited so‑called “pink slime” sites roughly half the time (ChatGPT >70%, Grok ~30%).
- Because many voters use chatbots for election information, widespread errors and partisan sourcing risk misinforming voters and undermining electoral integrity.
WHY IT MATTERS
Because many voters use chatbots for election information, widespread errors and partisan sourcing risk misinforming voters and undermining electoral integrity.