A major new study has found that four of the most commonly used AI assistants, including ChatGPT, Microsoft's Copilot, Google's Gemini, and Perplexity AI, misrepresent news content a staggering 45% of the time. The research, conducted by 22 public service media organizations, including the BBC and NPR, evaluated over 2,700 responses from these AI chatbots and revealed some alarming errors.
The study highlights significant issues with sourcing, accuracy, and context, with Gemini performing the worst, having 76% of its responses flagged for sourcing issues. Factual errors were also prevalent, including Perplexity incorrectly stating that surrogacy is illegal in Czechia and ChatGPT erroneously identifying Pope Francis as the current Pope. These findings raise concerns about the reliability and potential for misinformation stemming from AI-powered news platforms.
The researchers are calling on tech companies to prioritize accuracy improvements and be more transparent about their results. Jean Philip De Tender, deputy director general of the European Broadcasting Union, emphasizes the importance of trust in news, stating, "When people don't know what to trust, they end up trusting nothing at all, and that can deter democratic participation". The study's authors stress that independent monitoring of AI assistants is crucial, given the rapid rollout of new AI models.
The implications of this study are far-reaching, and it's essential to address these issues to maintain public trust in news. As AI assistants become increasingly common for accessing information, it's crucial to ensure they provide accurate and reliable information. The study's findings serve as a wake-up call for tech companies to take responsibility for their products and prioritize accuracy and transparency.
