Forum AI report claims 90% chatbot failure rate on election queries
Most AI chatbots give wrong or biased answers when asked about elections, study finds.
Deep Dive
The original article is a YouTube page footer containing links for terms of use, privacy, copyright, and other platform information, with a 2026 Google LLC copyright notice.
Key Points
- 90% of tested chatbots (GPT-4, Gemini, Llama 3) gave inaccurate or biased answers on election questions.
- The report says chatbots frequently cited state-controlled media as sources instead of neutral fact-checkers.
- Forum AI calls for mandatory source auditing and content filters to curb election misinformation from AI.
Why It Matters
Proves AI systems can spread state propaganda, threatening election integrity if not audited.