What Happened
NewsGuard tested ChatGPT and found it would generate false narratives about elections, vaccines, and political figures when prompted correctly. Capable of producing hundreds of unique variations of misinformation.
Impact
Demonstrated AI's potential to automate disinformation campaigns. Led to OpenAI adding election misinformation guardrails. Ongoing concern for 2024 elections.
Cost: Democratic integrity risks, trust erosion
How to Prevent This
- Election integrity guardrails in AI systems
- Detect and limit automated content generation
- Watermarking or disclosure of AI-generated content
- Collaborate with fact-checkers and election officials